a new limit on lorentz- and cpt-violating neutron spin...
Post on 16-Aug-2020
1 Views
Preview:
TRANSCRIPT
A New Limit on
Lorentz- and CPT-Violating
Neutron Spin Interactions
Using a K-3He Comagnetometer
Justin Matthew Brown
A Dissertation
Presented to the Faculty
of Princeton University
in Candidacy for the Degree
of Doctor of Philosophy
Recommended for Acceptance
by the Department of
Physics
Adviser: Michael V. Romalis
September 2011
c© Copyright by Justin Matthew Brown, 2011.
All rights reserved.
Abstract
Gravity and quantum mechanics are expected to unify at the Planck scale described
by an exceedingly large energy of 1019 GeV. This regime is far from the reach of the
highest energy colliders, but tests of fundamental symmetries provide an avenue to
explore physics at this scale from the low energy world. Proposed theories of quantum
gravity suggest possible spontaneous breaking of Lorentz and CPT symmetry that
have so far been unobserved. This thesis presents a test of a Lorentz- and CPT-
violating background field constant across the solar system coupling to the neutron
spin. A comagnetometer with polarized atomic vapors K and 3He suppresses magnetic
fields but remains sensitive to non-magnetic spin couplings. In a tabletop setup,
the comagnetometer measures the equatorial components to be bnX = (0.1 ± 1.6) ×
10−33 GeV and bnY = (2.5±1.6)×10−33 GeV, improving the previous neutron coupling
limit by a factor of 30.
This work utilizes the CPT-II apparatus, a second generation comagnetometer
installation which features a compact design and evacuated bell jar enclosing all optics
for improved long term stability. A rotating platform provides frequent reversals
of the apparatus orientation and represents a significant improvement in a Lorentz
violation search. The comagnetometer is also a sensitive gyroscope, so reorientation
with respect to Earth’s rotation contributes significantly to the signal. Reversals
along both the north-south and east-west directions provide a separation of systematic
effects from maximal and minimal gyroscope backgrounds. Systematic background
effects associated with reversals of the apparatus are also identified and removed.
Several novel features of comagnetometer behavior are also explored including the
response to an ac magnetic field and magnetic field gradients.
Furthermore, a compact, nuclear spin source containing 9× 1021 hyper-polarized
3He spins is designed and constructed. The spin source is used in conjunction with
the first generation apparatus CPT-I for a test of proposed long-range spin-dependant
iii
forces on the scale of 50 cm. The spin source supports accurate real-time monitoring
of the 3He polarization, an active magnetic field cancelation system, and efficient spin
reversals with losses below 2.5×10−6 per flip. This experiment leads to a new limit on
neutron coupling to light pseudoscalar and vector particles, including torsion, along
with constraints on possible couplings in recently proposed models involving unpar-
ticles and spontaneous breaking of Lorentz symmetry. This measurement improves
the previous limit by a factor of 500 and reaches an energy resolution of 10−34 GeV,
the highest energy resolution of any atomic experiment.
iv
Acknowledgements
The work in this thesis represents long hours toiling in the basement of Jadwin Hall
far from sunlight and regular human interaction associated with the outside world. I
couldn’t have done this work without the help and support of many colleagues and
friends. I would like to acknowledge a few of them here.
I would like to thank my adviser Michael Romalis for the exceptional choice of
research projects and teaching me not to be afraid to build it myself. I’ve certainly
admired the unceasing drive with which he pursues science and that he never gives
up because there is always something else to try. I look back on the work that I’ve
completed in the lab, and I’m amazed with the pace and the quantity of things that
I’ve managed to learn. Mike is one of a handful of people who have pushed me harder
than I would have pushed myself, but at the same time has been fair, particularly
with shoulder surgery near the end of my degree. I appreciate the potential that
he has seen in me and the encouragement to apply for academic fellowships. I’m
looking forward to continuing my scientific career and certainly wouldn’t have my
next opportunity without his support and guidance. I’d also like to thank William
Happer for his constructive comments provided throughout this research and serving
as my second reader on short notice. I’ve also enjoyed the opportunity to participate
in the annual Happer Lab canoe trip to the Pine Barrens each fall.
There are many individuals within the physics department and university staff who
deserve recognition for their time spent helping and inspiring me. Tsering (Wangyal)
Shawa in the Geoscience library was key to tracking down the orthorectified aerial
photo of Jadwin Hall and confirming which direction is True North. Jane Holmquist
in the Lewis Library has been my number one resource for obscure research papers
and books. I’d especially like to thank the Physics Department Purchasing team,
Claudin Champagne, Barbara Grunwerg, and Mary Santay, for their efficiency at
handling all of our many purchasing requests and for inviting me into their office to
v
chat even when they were busy. Mike Pelso and Ted Lewis have been instrumental in
helping me machine projects in a timely and safe manner while Bill Dix, Fred Norton,
and Glenn Atkinson machined items far too complicated for me to do myself. Mike
Souza hand blew the high pressure glass cell that survived the entire spin-dependant
forces measurement without exploding. Regina Savadge never failed to point out an
opportunity to scrounge for leftover food as well as provide a friendly comment to
keep me grounded in reality. Omelan Stryzak has been an ally for all sorts of random
technical advice relating to the lab or my hobbies. Finally, Laurel Lerner has been a
great friend and has served as my “Jewish Mother”1.
There are many colleagues in the Romalis group who have made my time in the
basement more pleasant: Hoan Dang, Giorgos Vasilakis, Scott Seltzer, Rajat Ghosh,
Sylvia Smullin, Tom Kornack, Vishal Shah, Lawrence Cheuk, Pranjal Vachaspati,
Katia Mehnert, Nezih Dural, Marc Smiciklas, Junhui Shi, and Shuguang Li. Tom
Kornack and Sylvia Smullin laid the groundwork for the Lorentz violation measure-
ment and I appreciate the opportunity to stand on their shoulders to finish what
they started. I would like to thank Scott Seltzer (aka “The Great Scott”) for two
inspiring works of great science in Refs. [1, 2]. One day, I hope that the moderators
of Wikipedia reach a level of competency such that they recognize the significance
of these works and accept them as proper citations2. I would also like to thank my
friend and colleague Giorgos Vasilakis for always being willing to lend a hand and for
our many physics discussions and arguments. I’ve learned much more beyond this
research in weekly journal club meetings with Ben Olsen and Bart McGuyer. I admire
Ben’s seemingly effortless way of synthesizing a complex idea into something simple
and that he always has a question about some new topic. Bart is truly a guy you want
on your team if you want to get something done and an inspiration for “extra-research
1Don’t worry, Mom, she hasn’t actually been able to adopt me, though she certainly has tried.2Can you believe that they questioned his/their notability?
vi
projects”3. It’s an incredible honor to be considered Bob Austin’s stereotypical nerd.
Mike Schroer, Mike Kolodrubitz, Richard Saldanha, Ben (and Rachel) Loer, Ryan
(and Julie) Fisher, Cynthia Chiang, Sasha Rahlin, David Liao and others have been
great friends around the department. Where would any of us in physics be without a
free T-shirt or box of labsnacks from Thorlabs? The personal email from Alex Cable,
founder of Thorlabs, regarding the missing labsnacks was certainly a highlight of my
career.
Many friends have provided great reasons to find time away from the lab. I’ll
never forget the many creative diversions from research with Jeff Dwoskin Ph.D.,
Sam Taylor, Jess Hawthorne, Matt Meola, Ana Pop, CJ Bell, and Seth Dorfman in-
cluding Formal Thursday, The Princeton Gingerbread Museum4, balloon pits, pillow
fights, Kidnap the Bride, Fish Fest, surprise weddings etc. It has been an honor and a
privilege to play with Ben Jorns and CJ Bell in the Action Brass Ensemble, Blawen-
berg Community Band, Physics Department Recital5, and the Princeton University
Wind Ensemble (PUWE). I am grateful for the recent leadership of Bob Gravener
and George Che who have brought PUWE to a college level performance group and
have given me a regular place to unwind at the end of a long day in the lab. I am
impressed by John MacDonald, Wilber Stewart, and Sharif Sazzad: community mem-
bers who took a technical interest in my work and whose eyes never glazed over when
I described my projects. I continue to look forward to Williams College Marching
Band Alumni gatherings with Cathy (Bryant) Van Orden, Charlie and Lida Doret,
Art and Katy Munson, Kate Alexander, Steve Wollkind, Josh and Rachelle Ain, Seth
Brown, and Patchen Mortimer.
3For instance: B. H. McGuyer, J. M. Brown, and H. B. Dang. Diet Soda and Liquid Nitrogen.American Journal of Physics 77, 8 pp. 667 (2009).
4It may be closed but will live on in memory. For the record, The Princeton Gingerbread Museumis not a legitimate business nor is it affiliated with Princeton University in any way.
5One of the main reasons that I chose the Princeton program above other competitive offers.
vii
I have been fortunate to have many extended family members nearby for family
gatherings, and I treasure the frequent visits with my grandparents Richard and Marie
Drake. I’ve also enjoyed the opportunity to see Ken and Donna Drake, Susan and
Lenny Cutugno, Richard and Nadine Drake, Robert and Ann Drake, Stephen and
Leslie Drake, and my many cousins. The negotiation skills learned from Ken Drake
and civil engineering insight6 from Kent Brown have been invaluable to me in this
work.
While not always nearby, my sister Alison (Brown) Cheung and brother Scott
Brown always know how to give me a hard time, but they are two of the few individuals
who appreciate what this document really represents, that they will be related to
“Doc Brown”. Last but not least, I would like to thank my parents, Garth and
Elaine Brown. I never would have dreamed of completing projects like this without
their support, guidance, and encouragement to always “do my best”. It was always
clear that science, organization, and solving problems was in my blood and where I
inherit these traits from. Thank you just doesn’t say enough.
6As well as the free Brown Engineering shirt.
viii
In memory of Richard H. Drake
ix
Contents
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xviii
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xix
1 Introduction 1
1.1 Lorentz and CPT Violation . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.1 Lorentz Symmetry . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.2 CPT Symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Lorentz Violation Searches . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.1 Standard Model Extension . . . . . . . . . . . . . . . . . . . . 6
1.2.2 Nuclear Spin Anisotropy Searches . . . . . . . . . . . . . . . . 8
1.3 Dissertation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3.1 K-3He Comagnetometer . . . . . . . . . . . . . . . . . . . . . 11
1.3.2 Unit Conventions . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.3.3 Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2 Background 14
2.1 Alkali-Metal Vapor . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.1.1 Energy Levels . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1.2 Light Absorption . . . . . . . . . . . . . . . . . . . . . . . . . 15
x
2.1.3 Lightshifts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.1.4 Optical Pumping . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2 Spin-Exchange Optical Pumping . . . . . . . . . . . . . . . . . . . . . 22
2.2.1 Spin-Exchange Collisions . . . . . . . . . . . . . . . . . . . . . 22
2.2.2 Longitudinal Relaxation: T1 . . . . . . . . . . . . . . . . . . . 24
2.3 Larmor Precession . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.3.1 Isolated Spin . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.3.2 Bound Electron . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.4 Alkali-Metal SERF Magnetometer . . . . . . . . . . . . . . . . . . . . 30
2.4.1 Magnetometry . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.4.2 SERF Regime . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.4.3 SERF Magnetometer . . . . . . . . . . . . . . . . . . . . . . . 34
2.5 Interacting Spin Ensembles . . . . . . . . . . . . . . . . . . . . . . . . 35
2.5.1 Contact Interaction . . . . . . . . . . . . . . . . . . . . . . . . 35
2.5.2 Coupled K-3He Dynamics . . . . . . . . . . . . . . . . . . . . 38
2.5.3 Oscillatory Magnetic Field Suppression . . . . . . . . . . . . . 42
3 Next Generation K-3He Comagnetometer 44
3.1 Comagnetometer Overview . . . . . . . . . . . . . . . . . . . . . . . . 44
3.1.1 Comagnetometer Description . . . . . . . . . . . . . . . . . . 45
3.1.2 Practical Anomalous Fields . . . . . . . . . . . . . . . . . . . 48
3.1.3 Complete Bloch Equations . . . . . . . . . . . . . . . . . . . . 50
3.2 Steady State Response . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.2.1 Leading Response . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.2.2 Numerator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.2.3 Denominator . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.2.4 Anomalous Fields . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.2.5 Pumping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
xi
3.2.6 Magnetic Fields . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.2.7 Rotations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.2.8 Lightshifts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.2.9 Refinements . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.2.10 Typical Leading Terms . . . . . . . . . . . . . . . . . . . . . . 64
3.3 Experimental Implementation . . . . . . . . . . . . . . . . . . . . . . 64
3.3.1 General Features . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.3.2 Historical Development . . . . . . . . . . . . . . . . . . . . . . 66
3.4 CPT-II Apparatus . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.4.1 CPT-II Design . . . . . . . . . . . . . . . . . . . . . . . . . . 76
3.4.2 CPT-II Performance . . . . . . . . . . . . . . . . . . . . . . . 87
3.4.3 An Improved Lorentz Violation Search . . . . . . . . . . . . . 91
4 Rotating Comagnetometer Measurements 95
4.1 Reversal Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . 95
4.1.1 Rotation Conventions . . . . . . . . . . . . . . . . . . . . . . . 96
4.1.2 Gyroscope Measurements . . . . . . . . . . . . . . . . . . . . 97
4.1.3 Optical Rotation Measurements . . . . . . . . . . . . . . . . . 98
4.1.4 Tilt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
4.1.5 Faraday Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
4.1.6 Optical Interference . . . . . . . . . . . . . . . . . . . . . . . . 105
4.1.7 Optical Rotation Background Measurements . . . . . . . . . . 107
4.2 Comagnetometer Zeroing . . . . . . . . . . . . . . . . . . . . . . . . . 109
4.2.1 Quasi-Steady State Measurements . . . . . . . . . . . . . . . . 109
4.2.2 Compensation Point . . . . . . . . . . . . . . . . . . . . . . . 110
4.2.3 Zero Bz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
4.2.4 Zero By . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
4.2.5 Zero Bx . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
xii
4.2.6 Orientation Dependent Zeroing Shifts . . . . . . . . . . . . . . 113
4.2.7 Zero Lx and ~sm . . . . . . . . . . . . . . . . . . . . . . . . . . 115
4.2.8 Zero Ly and Lz . . . . . . . . . . . . . . . . . . . . . . . . . . 116
4.2.9 Zero Pump-Probe Orthogonality . . . . . . . . . . . . . . . . . 117
4.2.10 Zeroing Convergence . . . . . . . . . . . . . . . . . . . . . . . 119
4.3 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
4.3.1 Traditional Calibration: ByBz Slope . . . . . . . . . . . . . . 120
4.3.2 Ωy Rotations . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
4.3.3 Ωx Sensitivity Through Zeroing . . . . . . . . . . . . . . . . . 123
4.3.4 Ωz Rotations . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
4.3.5 Low Frequency Bx Modulation . . . . . . . . . . . . . . . . . 128
4.3.6 Traditional Calibration Revisited . . . . . . . . . . . . . . . . 131
4.4 Gyroscope Phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
4.4.1 Phase Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . 134
4.4.2 Lightshift Feedback . . . . . . . . . . . . . . . . . . . . . . . . 138
4.5 AC Response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
4.5.1 Low Frequency Response . . . . . . . . . . . . . . . . . . . . . 142
4.5.2 AC Heater Coupling . . . . . . . . . . . . . . . . . . . . . . . 144
4.5.3 Sensitivity Optimization . . . . . . . . . . . . . . . . . . . . . 147
4.6 Symptomatic/Enhanced 3He Relaxation . . . . . . . . . . . . . . . . 147
4.6.1 Self-Relaxation . . . . . . . . . . . . . . . . . . . . . . . . . . 148
4.6.2 Magnetized Comagnetometer Cell . . . . . . . . . . . . . . . . 150
5 A Test of Lorentz and CPT Violation 154
5.1 Lorentz Violation Search . . . . . . . . . . . . . . . . . . . . . . . . . 154
5.1.1 Measurement Protocol . . . . . . . . . . . . . . . . . . . . . . 155
5.1.2 Zeroing Schedule . . . . . . . . . . . . . . . . . . . . . . . . . 157
5.1.3 Reliability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
xiii
5.1.4 Least Squares Fits . . . . . . . . . . . . . . . . . . . . . . . . 162
5.1.5 Combined Measurements . . . . . . . . . . . . . . . . . . . . . 163
5.2 Correlation Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
5.2.1 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
5.2.2 Faraday Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
5.2.3 Pump Lightshift . . . . . . . . . . . . . . . . . . . . . . . . . 166
5.2.4 Bz Drifts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
5.2.5 Orientation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
5.2.6 Long Term Signal Variations . . . . . . . . . . . . . . . . . . . 172
5.3 Analysis Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
5.3.1 Final Processing Scheme . . . . . . . . . . . . . . . . . . . . . 174
5.3.2 Processing Methods . . . . . . . . . . . . . . . . . . . . . . . . 175
5.3.3 Calibrated Data . . . . . . . . . . . . . . . . . . . . . . . . . . 178
5.3.4 Domains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
5.3.5 Averaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
5.3.6 Least Squares Fit Refinements . . . . . . . . . . . . . . . . . . 180
5.3.7 Systematic Checks . . . . . . . . . . . . . . . . . . . . . . . . 183
5.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
5.4.1 Representative File and Systematic Uncertainty . . . . . . . . 184
5.4.2 Sidereal Amplitudes . . . . . . . . . . . . . . . . . . . . . . . 184
5.4.3 Final Average . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
5.4.4 SME Parameters and Energy Scaling . . . . . . . . . . . . . . 189
5.4.5 Interpretation to a Limit . . . . . . . . . . . . . . . . . . . . . 191
5.4.6 Competitors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
5.4.7 Beyond bn⊥ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
5.4.8 Proposed Improvements . . . . . . . . . . . . . . . . . . . . . 196
xiv
6 Nuclear Spin Source 199
6.1 Spin Source Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
6.1.1 Early Spin Source Considerations . . . . . . . . . . . . . . . . 200
6.1.2 Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
6.1.3 Design Considerations . . . . . . . . . . . . . . . . . . . . . . 202
6.1.4 Cell Manufacture . . . . . . . . . . . . . . . . . . . . . . . . . 204
6.2 Efficient Spin Reversals . . . . . . . . . . . . . . . . . . . . . . . . . . 206
6.2.1 Adiabatic Fast Passage . . . . . . . . . . . . . . . . . . . . . . 206
6.2.2 Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
6.2.3 Magnetic Field Optimization . . . . . . . . . . . . . . . . . . . 208
6.2.4 Performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
6.3 3He Polarimetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
6.3.1 Electron Parametric Resonance . . . . . . . . . . . . . . . . . 212
6.3.2 Polarization Measurement . . . . . . . . . . . . . . . . . . . . 212
6.4 Magnetic Field Compensation . . . . . . . . . . . . . . . . . . . . . . 215
6.4.1 Compensation Coil . . . . . . . . . . . . . . . . . . . . . . . . 215
6.4.2 EPR Consideration . . . . . . . . . . . . . . . . . . . . . . . . 215
6.4.3 Performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
6.5 Spin Source Operation . . . . . . . . . . . . . . . . . . . . . . . . . . 216
6.5.1 Flipping Sequence . . . . . . . . . . . . . . . . . . . . . . . . . 217
6.5.2 Long-Range Spin-Dependant Forces Search . . . . . . . . . . . 218
6.5.3 Future Improvements . . . . . . . . . . . . . . . . . . . . . . . 220
7 Conclusion 221
A Time Conventions 223
B Local Coordinate System 226
B.1 Laboratory Location . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
xv
B.1.1 Jadwin Hall Exterior . . . . . . . . . . . . . . . . . . . . . . . 228
B.1.2 Jadwin Hall Interior . . . . . . . . . . . . . . . . . . . . . . . 232
B.1.3 Comagnetometer Orientation . . . . . . . . . . . . . . . . . . 232
B.1.4 Magnetic North . . . . . . . . . . . . . . . . . . . . . . . . . . 234
B.2 Coordinate Frame Variations . . . . . . . . . . . . . . . . . . . . . . . 236
B.2.1 Polar Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
B.2.2 Length of Day . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
B.2.3 Earthquakes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
B.2.4 Comment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
B.3 Rotation Conventions . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
C Comagnetometer Toy Model 241
C.1 Steady State Compensation . . . . . . . . . . . . . . . . . . . . . . . 241
C.1.1 Derivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
C.1.2 Steady State Equation Lists . . . . . . . . . . . . . . . . . . . 245
C.1.3 Primary Magnetic Field Dependence . . . . . . . . . . . . . . 246
C.1.4 Products of Magnetic Fields . . . . . . . . . . . . . . . . . . . 247
C.1.5 Rotations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
C.1.6 Lightshifts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
C.1.7 Anomalous Fields . . . . . . . . . . . . . . . . . . . . . . . . . 252
C.1.8 Signal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
D Lorentz Violation Summary 255
E Signal Analysis 261
E.1 Uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
E.1.1 Independent Measurements . . . . . . . . . . . . . . . . . . . 262
E.1.2 Weighted Average and Uncertainty . . . . . . . . . . . . . . . 262
E.1.3 Reduced χ2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
xvi
E.1.4 Error Propagation . . . . . . . . . . . . . . . . . . . . . . . . 264
E.2 Over-Sampled Measurements . . . . . . . . . . . . . . . . . . . . . . . 265
E.3 String Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266
E.3.1 String Analysis Refinement . . . . . . . . . . . . . . . . . . . 268
Bibliography 271
xvii
List of Tables
1.1 C, P, and T Symmetries . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2 Historical Nuclear Spin Anisotropy Searches . . . . . . . . . . . . . . 10
3.1 Bloch Equation Notation . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.2 Estimated Comagnetometer Parameters . . . . . . . . . . . . . . . . 54
3.3 Comagnetometer Spin-Exchange Contributions . . . . . . . . . . . . . 56
3.4 Motor Speeds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.1 Verdet Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
4.2 Zeroing Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
4.3 Ωz Sensitivity Measurements . . . . . . . . . . . . . . . . . . . . . . . 127
4.4 Heater Panels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
5.1 Encoder Variations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
5.2 Representative Amplitude Summary . . . . . . . . . . . . . . . . . . 188
5.3 Best b⊥ Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . 196
6.1 Spin Source Coil Calibrations . . . . . . . . . . . . . . . . . . . . . . 209
D.1 Lorentz Violation Sampling Intervals . . . . . . . . . . . . . . . . . . 258
D.2 Lorentz Violation Datasets Overview . . . . . . . . . . . . . . . . . . 259
D.3 Lorentz Violation Dataset Rotation Parameters . . . . . . . . . . . . 260
xviii
List of Figures
1.1 Charge, Parity, and Time Reversals . . . . . . . . . . . . . . . . . . . 6
1.2 Proposed Lorentz-Violating Field . . . . . . . . . . . . . . . . . . . . 9
1.3 Historical Nuclear Spin Anisotropy Searches . . . . . . . . . . . . . . 10
2.1 Energy Levels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2 Optical Pumping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.3 Spin-Exchange Optical Pumping . . . . . . . . . . . . . . . . . . . . . 23
2.4 Larmor Precession . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.5 Magnetometer Scheme . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.6 Precession in the SERF Regime . . . . . . . . . . . . . . . . . . . . . 32
2.7 Fast Damping at the Compensation Point . . . . . . . . . . . . . . . 40
3.1 Comagnetometer Description . . . . . . . . . . . . . . . . . . . . . . 47
3.2 Comagnetometer Gyroscope . . . . . . . . . . . . . . . . . . . . . . . 49
3.3 General Comagnetometer Apparatus . . . . . . . . . . . . . . . . . . 65
3.4 CPT-I Comagnetometer Installation . . . . . . . . . . . . . . . . . . . 68
3.5 CPT-I Lorentz Violation Results . . . . . . . . . . . . . . . . . . . . 71
3.6 CPT-I Upgrade . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
3.7 CPT-I Upgrade Side View . . . . . . . . . . . . . . . . . . . . . . . . 74
3.8 Best CPT-I Noise Spectrum . . . . . . . . . . . . . . . . . . . . . . . 75
3.9 CPT-II Schematic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
xix
3.10 Cell, Oven, and Shields . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.11 CPT-II Optical Layout . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.12 Beam Motion Spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
3.13 CPT-II Image . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
3.14 Four Faces of CPT-II . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
3.15 CPT-II Noise Spectrum . . . . . . . . . . . . . . . . . . . . . . . . . 89
3.16 Earth’s Rotation Response . . . . . . . . . . . . . . . . . . . . . . . . 92
3.17 Long Term Measurement . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.1 Typical Rotation Interval . . . . . . . . . . . . . . . . . . . . . . . . . 98
4.2 Polarimeter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.3 Photodiode Response . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
4.4 Beam Positions Under Reversals . . . . . . . . . . . . . . . . . . . . . 102
4.5 Optical Rotation from Tilt . . . . . . . . . . . . . . . . . . . . . . . . 103
4.6 Faraday Rotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
4.7 Probe Interference . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
4.8 Reversals with Background Measurements . . . . . . . . . . . . . . . 107
4.9 Long Term Background Removal . . . . . . . . . . . . . . . . . . . . 108
4.10 Quasi-Steady State Measurements . . . . . . . . . . . . . . . . . . . . 110
4.11 Bz Modulation Example . . . . . . . . . . . . . . . . . . . . . . . . . 112
4.12 Orientation Dependant Zeroing . . . . . . . . . . . . . . . . . . . . . 114
4.13 Bz Resonance Map . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
4.14 Ω⊕y Calibration Response . . . . . . . . . . . . . . . . . . . . . . . . . 122
4.15 By Orientation Dependant Zeroing Calibration . . . . . . . . . . . . . 124
4.16 Signal Response to Ωz Rotations and By Detuning . . . . . . . . . . . 125
4.17 Signal Shifts to Ωz Rotations vs. By Detuning . . . . . . . . . . . . . 126
4.18 Low Frequency Bx Calibration . . . . . . . . . . . . . . . . . . . . . . 129
4.19 dBz/dz Resonance Broadening . . . . . . . . . . . . . . . . . . . . . . 132
xx
4.20 Gyroscope Calibration Comparison . . . . . . . . . . . . . . . . . . . 133
4.21 Gyroscope Phase Shift - Bz . . . . . . . . . . . . . . . . . . . . . . . 135
4.22 Bz Gyroscope Shift Summary . . . . . . . . . . . . . . . . . . . . . . 136
4.23 Lz Gyroscope Shift Summary . . . . . . . . . . . . . . . . . . . . . . 137
4.24 Bz Resonance Asymmetry for Lz . . . . . . . . . . . . . . . . . . . . 139
4.25 Lightshift Beam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
4.26 Low Frequency Response . . . . . . . . . . . . . . . . . . . . . . . . . 143
4.27 Heater Coupling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
4.28 3He Self-Relaxation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
4.29 Magnetized Cell: T1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
4.30 Low-field T1 Hysteresis . . . . . . . . . . . . . . . . . . . . . . . . . . 152
5.1 Complete Dataset . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
5.2 3He Polarization Drifts . . . . . . . . . . . . . . . . . . . . . . . . . . 159
5.3 Long Term Measurement and Fit . . . . . . . . . . . . . . . . . . . . 163
5.4 Calibration Comparison . . . . . . . . . . . . . . . . . . . . . . . . . 165
5.5 Calibrated Signal Comparisons . . . . . . . . . . . . . . . . . . . . . 166
5.6 Long Term External Magnetic Field Changes . . . . . . . . . . . . . . 167
5.7 Pump Lightshift Correction . . . . . . . . . . . . . . . . . . . . . . . 168
5.8 Bz Drift Correction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
5.9 Platform Position . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
5.10 Vibration Isolation Stage Position . . . . . . . . . . . . . . . . . . . . 172
5.11 Fourier Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
5.12 Largest Drift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
5.13 Processing Modification Example . . . . . . . . . . . . . . . . . . . . 176
5.14 Gyroscope Amplitude Summary . . . . . . . . . . . . . . . . . . . . . 178
5.15 Strong Fit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
5.16 Representative File . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
xxi
5.17 Final Amplitudes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
5.18 Simulated Lorentz-Violating Field . . . . . . . . . . . . . . . . . . . . 187
5.19 Gaussian Interpretation . . . . . . . . . . . . . . . . . . . . . . . . . 191
5.20 Recent bn⊥ Limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
6.1 Comagnetometer with Spin Source . . . . . . . . . . . . . . . . . . . 200
6.2 Spin Source Schematic . . . . . . . . . . . . . . . . . . . . . . . . . . 202
6.3 Spin Source Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
6.4 Magnetic Field Coil Design . . . . . . . . . . . . . . . . . . . . . . . . 211
6.5 EPR Frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
6.6 EPR Frequency Under Reversals . . . . . . . . . . . . . . . . . . . . . 214
6.7 Compensation Coil Reversals . . . . . . . . . . . . . . . . . . . . . . 216
6.8 Long-Range Spin-Dependant Forces Measurement . . . . . . . . . . . 219
A.1 Celestial Coordinate System . . . . . . . . . . . . . . . . . . . . . . . 224
B.1 Overlap with Earth’s Rotation . . . . . . . . . . . . . . . . . . . . . . 227
B.2 Aerial View of Jadwin Hall . . . . . . . . . . . . . . . . . . . . . . . . 229
B.3 Jadwin Hall Survey . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
B.4 Jadwin Hall Basement Floor Plan . . . . . . . . . . . . . . . . . . . . 233
B.5 CPT-II Lab Layout . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
B.6 Length of Day and Polar Motion Variations . . . . . . . . . . . . . . 237
C.1 3He Cancelation Imperfection . . . . . . . . . . . . . . . . . . . . . . 242
E.1 Effective Noise Bandwidth . . . . . . . . . . . . . . . . . . . . . . . . 266
xxii
Chapter 1
Introduction
The Standard Model of Particle Physics unifies three of the four fundamental forces,
the electromagnetic, the weak, and strong into one model using continuous rotations
of groups SU (3) × SU (2) × U(1). General relativity describes gravity, space, and
time on a grand scale through geometry. These two theories span the full scale of
possible interactions observed in nature from quantum mechanics to the formation of
the universe. Each theory has enormous predictive power, yet the two have not been
reconciled. Attempts to unify these forces in a single theory of quantum gravity or
Grand Unified Theory have led to widely varying models that have yet to be verified
with any experimental signature.
It is expected that General Relativity and the Standard Model are low energy
limits of a more fundamental theory. A simple scaling argument combines the fun-
damental constants from quantum mechanics ~, space-time c, and gravity G into
a complete set of characteristic length, time, and mass scales known as the Planck
units. No combination of these constants can produce a dimensionless parameter, so
these units are dimensionally independent [3]. The Planck mass is defined as
MP =
√~cG
= 1.2× 1019 GeV/c2 (1.1)
1
and is expected to be a boundary between standard theories of gravity and quantum
mechanics.
Two mainstream theories compete to describe Planck scale physics, Loop Quan-
tum Gravity [4] and String Theory [5]. String theory unifies gravity and gauge fields
into a consistent quantum theory, but does not predict a 4-D full Standard Model
phenomenology without unbroken supersymmetry. Loop Quantum Gravity provides
an alternative unification of quantum mechanics and gravity but does not aim to
describe all interactions as does String Theory. These theories distinctly differ in
the treatment of spacetime as a fundamental or background field. A comprehensive
comparison of these theories is beyond the scope of this work, but it suffices to point
out that there have been no observed experimental signatures of either theory. Both
are considered tentative and require experimental validation.
At present, physics at the highest energy scale is directly probed by the Large
Hadron Collider (LHC). The LHC collides two proton beams at an energy of
7 TeV/beam. These events represent the highest energies ever attained in the labora-
tory under controlled conditions, but are still far from probing physics at or near the
Planck scale. Astrophysical tests can explore these high energies. Hawking radiation
is emitted from a black hole if it has a Hawking temperature above the temperature
of the cosmic microwave background radiation. In the final stages of evaporation,
the Hawking radiation should approach the Planck energy, but this radiation has yet
to be observed [6]. Similarly, propagation of energy from distant gamma ray bursts
can be affected by the Planck scale granularity of spacetime [7]. It is also expected
that an imprint of primordial gravitational radiation can exist within the cosmic
microwave background [8]. Conditions in these experiments are impossible to control
and repeat under different parameters. If new physics is present at the level of the
Planck scale, it is expected that experiments of exceptional sensitivity may provide
a signature of these effects [9].
2
1.1 Lorentz and CPT Violation
As an alternative to high energy approaches, precision laboratory scale tests provide a
unique opportunity to probe the highest energy scales within a controlled environment
through tests of fundamental symmetries. Observable signals from a fundamental
theory may also reveal themselves at a more accessible level, perhaps only suppressed
by the ratio of the electroweak mW and Planck scales where mW/MP ' 10−17 [9].
Lorentz and CPT symmetry are fundamental tenets of both the Standard Model
and General Relativity, so evidence of a broken symmetry immediately provides a
signature of new physics. Any realistic Lorentz-covariant fundamental theory with
more than 4 dimensions is expected to produce spontaneous Lorentz violation [10],
and it has already been shown that String Theory can lead to Lorentz and CPT
violation [11]. It has also been shown that a feature of an interacting quantum field
theory is that Lorentz violation will follow from CPT violation [12]. No precision
measurement has yet to observe an experimental signature of quantum gravity.
1.1.1 Lorentz Symmetry
Lorentz symmetry states that the laws of physics should be the same in all inertial
frames. This statement is expressed mathematically in terms of a Lorentz transforma-
tion. For a co-ordinate 4-vector xµ = (ct, x, y, z), an example Lorentz transformation
is the Lorentz boost where xµ transforms to x′µ along the x-axis with a velocity v as
t′ =βγ(t− vx/c2)
x′ =βγ(x− vt)
y′ =y
z′ =z (1.2)
3
where βγ = 1/√
1− (v/c)2. This transformation preserves the metric of flat spacetime
ds2 = −(cdt)2 + dx 2 + dy2 + dz 2 (1.3)
as an invariant. In general, a Lorentz transformation is any transformation that
preserves the metric of flat spacetime (Eq. 1.3) and is called covariant [13].
Lorentz transformations can be broken into two types, observer Lorentz transfor-
mations and particle Lorentz transformations. An observer Lorentz transformation
is described by
x′µ = Λνµxν + aµ (1.4)
where Λνµ describes rotations and boosts while aµ describes translations. These trans-
formations describe observations made in two inertial frames with differing orienta-
tions and velocities. A particle Lorentz transformation for a particle φ follows
U(a,Λ)φ(x)U(a,Λ)† = φ(Λx+ a) (1.5)
where U describes all of the boosts, rotations, and translations. These transforma-
tions relate the properties of two particles with different spin or momentum within
a specific oriented inertial frame. In the absence of Lorentz violation, these trans-
formations are equivalent, but in a Lorentz-violating theory, particles at different
velocities experience different physics. Therefore, particle Lorentz transformations
lead to experimentally observable deviations from the predicted theory.
1.1.2 CPT Symmetry
In contrast to Lorentz symmetry, CPT symmetry is a discrete symmetry under charge
conjugation C, parity inversion P, and time reversal T. Effectively, charge conjugation
casts a particle to its corresponding anti-particle [14], parity inversion considers a
4
“mirror” world, and time reversal switches a particle moving forward in time to one
moving backwards in time (Fig. 1.1). Examples of these transformations on various
physical parameters are listed in Table 1.1.
The CPT Theorem states that in a relativistic local quantum field theory the
combination of charge, parity, and time will be good symmetries under the conditions
of Lorentz invariance, hermiticity of the Hamiltonian, and locality of the theory [15,
16]. A violation of CPT symmetry implies that one or more of these assumptions is
false. The Standard Model predicts CPT as a perfect symmetry, so any evidence of
CPT violation is evidence of new physics.
Typical CPT tests involve comparison of the mass, lifetime, and the magnitude
of both the charge and magnetic moment of matter and antimatter. In general,
anti-matter experiments are especially challenging due to the difficult production
and isolation of anti-matter particles. Recent theoretical work has demonstrated a
connection between CPT and Lorentz violation and proposed tests of CPT without
antiparticles [12, 17]. A constraint on Lorentz violation provides a constraint on CPT
violation.
Many measurements as individual or pair symmetries of C, P , and T have been
shown to be violated in the weak interaction [18, 19, 20, 21], but so far, no experimen-
tal violation of the CPT combination has been observed. CPT violation is possible
in String Theory [11] and could lead to baryogenesis [22].
1.2 Lorentz Violation Searches
Tests of matter coupling to an anisotropy of space were first studied by Hughes
[24] and Drever [25] in considering the orientation of the spin of 7Li to a preferred
reference frame. This type of experiment is now referred to as a Hughes-Drever exper-
iment. Further experiments have improved upon the sensitivity of these early mea-
5
e- e+
z’
y’
x’
x
z
y
Charge Parity Time
Figure 1.1: Charge reversal changes a particle to its antiparticle. Parity reversal switcheseach of the coordinate axes. Note that the handedness has changed. Time reversal switchesthe direction of time. Each of these discrete symmetries is violated individually in theStandard Model, however, their product is preserved.
Variable C P TCoordinate ~x→ ~x ~x→ −~x ~x→ ~xTime t→ −t t→ t t→ −tMomentum ~p→ ~p ~p→ −~p ~p→ −~pEnergy ε→ ε ε→ ε ε→ ε
Angular Momentum ~L→ ~L ~L→ ~L ~L→ −~LSpin ~S → ~S ~S → ~S ~S → −~SCharge q → −q q → q q → q
Electric Field ~E → − ~E ~E → − ~E ~E → ~E
Magnetic Field ~B → − ~B ~B → ~B ~B → − ~B
Table 1.1: Behavior of standard quantities under the influence of charge conjugation C,parity inversion P, and time reversal T [23].
surements. A resurgence of interest in these searches has followed from the Lorentz-
and CPT-violating formalism of the Standard Model Extension (SME) developed by
Alan Kostelecky and colleagues [9, 10, 17]. This formalism provides an important
framework for identifying Lorentz- and CPT-violating interactions and connecting
theoretical predictions to experimental observations.
1.2.1 Standard Model Extension
The SME provides the most general framework for considering spontaneous Lorentz
violation at the level of the Standard Model. It considers all possible Lorentz-violating
6
terms based on the SU (3)× SU (2)×U(1) gauge structure by combining all Lorentz-
violating operators and controlling coefficients to form observer-invariant terms. Con-
sideration of all possible Lorentz-violating terms is challenging, so the minimal SME
restricts operators to mass dimension less than or equal to 4 [26]. The SME is a viable
framework in that it maintains energy-momentum conservation, observer Lorentz co-
variance, conventional quantization, hermiticity, microcausality, and power-counting
renormalizability [17]. Here, it is particle Lorentz violation that is spontaneously bro-
ken rather than observer Lorentz violation (Section 1.1.1). The advantage of the SME
is that it considers all possible mechanisms based on the Standard Model, and it can
then be used as a lexicon of possible Lorentz- and CPT-violating interactions where
identification of a particular interaction can validate or disprove classes of theories.
The entire SME is large and much like the Standard Model rarely written out
explicitly. A special case is often useful to consider, particularly in the context of
an experiment. For a spin-1/2 fermion ψ with mass m, the minimal SME includes
the usual Standard Model lagrangian and the following dimension 3- and 4-operators
that lead to Lorentz-violating terms [17]
L =1
2iψΓν
←→∂νψ − ψMψ (1.6)
where1
M = m+ aµγµ + bµγ5γ
µ +1
2Hµνσ
µν (1.7)
and
Γν = γν + cµνγµ + dµνγ5γ
µ + eν + ifνγ5 +1
2gλµνσ
λµ. (1.8)
The coefficients aµ, bµ, cµν , dµν , eµ, fµ, gλµν , and Hµν determine the magnitude of the
possible Lorentz-violating interactions and are real since L is hermitian. The coeffi-
cients in M have dimensions of mass and the coefficients in Γν are dimensionless. In
1The notation A←→∂ µ ≡ A∂µB − (∂µA)B.
7
addition, aµ, bµ, eµ, fµ, and gλµν are CPT-odd while cµν , dµν , and Hµν are CPT-even.
Reference [17] discusses a number of suppression factors, and each of the coefficients
have been shown to be small [27].
Equation 1.6 leads to modifications to the equation of motion as described in
Ref. [10]. Consideration of only nonzero aµ and bµ values produces the following
lagrangian,
L =1
2iψγµ
←→∂ µψ − aµψγµψ − bµψγ5γµ −mψψ. (1.9)
The variational principle yields
(iγµ∂µ − aµγµ − bµγ5γµ −m)ψ = 0 (1.10)
which reduces to the Dirac Equation for a free particle when aµ = bµ = 0. The con-
stant background fields aµ and bµ transform as two 4-vectors under observer Lorentz
transformations and as eight scalars under particle Lorentz transformations. Observer
Lorentz violation remains invariant while particle Lorentz violation is partly broken
and leads to observable deviations in the free particle trajectory. The SME assumes
that the Lorentz-violating fields are constant across the scale of the solar system, so
Earth-based searches rely on interactions that vary over the course of a sidereal day
(Fig. 1.2).
1.2.2 Nuclear Spin Anisotropy Searches
Several Hughes-Drever Lorentz violation searches have been performed over the years
using a variety of atomic species. These experiments search for an energy shift of
a spin compared to the local celestial frame. The SME predicts several types of
energy shifts that would appear as sidereal, semisidereal2, annual, and fixed-frame
changes in the measured energy. Comparing the limits on the SME coefficients across
2Second harmonic of sidereal oscillations.
8
Ω⊕
Sun
Earth
Lab
Figure 1.2: Proposed Lorentz-violating fields are constant across the scale of the solarsystem and can be observed from the Earth as an interaction marked by the length of asidereal day. Image Credit: [28].
different species is difficult due to the nuclear structure and geometrical factors in-
volved [17]. However, it is relatively straightforward to compare the limits in terms
of measured precession frequency. Figure 1.3 displays a timeline of improvements
to Lorentz violation limits for nuclear spins in terms of the approximate precession
frequency measured. The approximate measured values, species, and references are
listed in Table 1.2. In general, the limits on an anomalous neutron spin coupling
are consistently better than those for the proton, and a dramatic improvement in
the precision of the measurement occurs from the use of differential measurements
between two different spin species.
Prior to the work of this thesis, the most sensitive test was performed for the
neutron using the Harvard-Smithsonian 3He-129Xe dual noble-gas spin-maser [29].
This thesis uses a K-3He comagnetometer to reach an energy resolution3 of 0.7 nHz
to improve the previous limit by a factor of 30. This result was first reported in
Ref. [30] with further details described throughout this work.
3The limit is closer to 1 nHz as listed in Table 1.2.
9
1950 1960 1970 1980 1990 2000 2010 202010-12
10-10
10-8
10-6
10-4
10-2
100
102A
ppro
xim
ate
Fre
quen
cy P
reci
sion
(H
z)
Year
7Li
7Li
9Be+
199Hg-201Hg21Ne-3He
Cs-199Hg
129Xe/3He-maser
H-maser
K-3He
UCN
3He/129Xe
K-3He
133Cs
Figure 1.3: Summary of nuclear spin anisotropy searches in terms of the approximate spinfrequency precision.
Spin Species Frequency Author Reference Anisotropy
7Li 8 Hz Hughes et al. [24] Sid.7Li 40 mHz Drever [25] Sid.
9Be+ 100 µHz Prestage et al. [31] Sid./Semisid.199Hg-201Hg 1 µHz Lamoreaux et al. [32] Sid./Semisid.
21Ne-3He 1 µHz Chupp et al. [33] Semisid.Cs-199Hg 100 nHz Berglund et al. [34] Sid.H-maser 500 µHz Phillips et al. [35] Sid.
129Xe/3He-maser 50 nHz Bear et al. [29, 36] Sid.129Xe/3He-maser 150 nHz Cane et al. [37] Annual
K-3He 60 nHz Kornack [28] Sid.133Cs 100 µHz Wolf et al. [38] Sid./Semisid./Fixed
Ultracold neutron 10 µHz Altarev et al. [39] Sid.K-3He 1 nHz Brown et al. [30] Sid.
3He/129Xe 5 nHz Gemmel et al. [40] Sid.
Table 1.2: Summary of nuclear spin anisotropy searches in terms of the approximate spinfrequency precision. Entries are in order of publication date. Sidereal and semisiderealare abbreviated by Sid. and Semisid. respectively. In the case of multiple measurements, arepresentative value has been selected.
10
1.3 Dissertation
This dissertation describes the use of the K-3He comagnetometer to perform a Hughes-
Drever experiment to search for a Lorentz- and CPT-violating field coupling to the
nuclear spin of 3He. The result is interpreted in terms of the SME and sets a new
limit on b⊥ for the neutron that improves upon the previous limit by a factor of
30. This second generation experiment supersedes its first generation predecessor
by improving upon the long term stability of the apparatus using compact design
and evacuated bell jar to enclose all of the lasers and optics. In addition, the entire
apparatus is placed on a rotating platform for lock-in modulation of the proposed
Lorentz-violating field every 22 s. Throughout this experiment, many new aspects of
comagnetometer operation were also explored.
1.3.1 K-3He Comagnetometer
This thesis refers to the atomic species K and 3He in a K-3He comagnetometer. The
comagnetometer consists of two overlapping spin ensembles, electrons in a polarized
K vapor and nuclear spins in a hyper-polarized 3He buffer gas in a spherical vapor
cell. Where appropriate, such as spin-exchange optical pumping, these species will be
denoted (a)-alkali and (b)-noble. In the context of precision measurements and the
comagnetometer, we will refer to the electrons (e) and nuclear (n) species.
The K-3He comagnetometer suppresses magnetic interactions but remains sensi-
tive to non-magnetic spin couplings, specifically, the difference in coupling between
electron and nuclear spins. The behavior of these spin ensembles is unique in that
the strong coupling between the electrons and nuclear spins quickly damps precession
of the nuclear spins to the steady state within several seconds. The steady state re-
sponse provides a measurement of anomalous spin couplings, and the fast damping to
the steady state allows for frequent reversals of the apparatus. The comagnetometer
11
is also a sensitive gyroscope [41] which provides a large background response from the
rotation of the Earth and a complication in a Lorentz violation search.
1.3.2 Unit Conventions
The natural choice of units in this thesis are fT. Presentation of results in GeV,
eV, gauss, or Hz would also be acceptable, though historical development of the
K-3He comagnetometer from high sensitivity magnetometers leads to the choice of
fT. Readers dissatisfied with this selection can reference Sections 2.3.2 and 5.4.4 for
conversions to their preferred unit.
A Lorentz-violating field is characterized by a sidereal variation in the energy shift
of the spin. For convenience, long term measurements are recorded in sidereal days
from January 1, 2000, as is convention in the field. One sidereal day is the length
of time it takes the earth to rotate on its axis and precess in its orbit to reorient
to the same celestial reference point. This time is roughly 23 h 56 min. It is also
convention to distinguish between Local Sidereal Time (LST) and Greenwich Mean
Sidereal Time (GMST). This distinction depends on the location of the measurement
where LST differs from GMST in the relative longitude of the location of the mea-
surement to Greenwich, England. Technical details on these conventions are included
in Appendix A.
1.3.3 Structure
Chapter 2 provides a brief background of spin-exchange optical pumping, magnetome-
tery, and coupled spin-dynamics. These insights allow for a complete description of
the K-3He comagnetometer and the apparatus in Chapter 3. Two installations of
the K-3He comagnetometer have been used for two related experiments in this the-
sis. Development of our modern comagnetometer apparatus extends over ten years
with contributions from several students and post-doctoral researchers. The main
12
focus of this thesis is on the second generation apparatus to set a new limit on a
Lorentz-violating field in Chapter 5. This measurement represents the highest energy
resolution in any spin anisotropy measurement to date [30]. In performing this search,
many new features of the K-3He comagnetometer were discovered and are described
in Chapter 4. A search for a proposed long-range spin-dependant forces has been
performed using a hyper-polarized 3He spin source. The main details of this experi-
ment are in Refs. [42, 43] and will not be reported here. This measurement achieves a
frequency resolution of 18 pHz which represents the highest energy resolution of any
experiment. Specific features of the spin source including its construction and design
not included in these references are discussed in Chapter 6.
13
Chapter 2
Background
The K-3He comagnetometer described in Chapter 3 and the hyper-polarized 3He spin
source described in Chapter 6 consist of glass vapor cells containing a hot K vapor
and a high density 3He buffer gas. Circularly polarized light on resonance with the
D1 transition in K polarizes the alkali-metal vapor through optical pumping. Spin-
exchange collisions with the buffer gas transfer polarization to the nucleus of 3He
through the hyperfine interaction. This chapter provides a brief summary of the
relevant physics of these polarized spin ensembles. This is a well-developed field with
many readily available references. The interested reader is encouraged to explore
resources beyond this text for more details not included in the summary presented
here.
2.1 Alkali-Metal Vapor
Before considering the interaction of K and 3He, it is useful to consider the properties
of a K vapor in several amagats1 of 3He buffer gas (1020 atoms/cm3). Atomic K has
filled inner electron shells and a single valence electron. The hydrogen-like energy level
1An amagat is the a unit of number density equivalent to 1 atm of an ideal gas at 0C. Confusioncan often arise in defining this in terms of standard temperature and pressure. Ultimately, thenumber density of interest is 2.68× 1019/cm3.
14
4P
4S 4S1/2
4P3/2
4P1/2
D1
D2
770.1 nm
766.8 nm
Figure 2.1: The 4P energy levels are split into the 4P1/2 and 4P3/2 levels by spin-orbitcoupling. The 4S1/2 → 4P1/2 and 4S1/2 → 4P3/2 transitions are labeled D1 and D2 respec-tively.
structure simplifies treatment of the atomic structure. Collisions with the high density
3He buffer gas distort the electron wavefunction and further simplify discussion of the
relevant atomic structure.
2.1.1 Energy Levels
In the ground state, the outermost electron in K resides in the 4S1/2 state. The
atomic fine structure describes the splitting of the nearby 4P state into the 4P1/2 and
4P3/2 states from a combination of spin-orbit coupling and relativistic corrections
(Fig. 2.1). The spin-orbit interaction Hso = α~L · ~S creates eigenstates of total angular
momentum ~J = ~L + ~S. The hyperfine interaction Hhf = A~I · ~S further splits each
of these states ~F = ~I + ~J , but they are typically optically unresolved due to the
pressure broadening in several amagats of 3He buffer gas. The hyperfine splitting
will be relevant in Section 2.4.2 and 6.3.1 and addressed in context. In the absence
of possible pressure shifts from the 3He buffer gas, the D1 and D2 transitions are at
770.1 nm and 766.8 nm respectively.
2.1.2 Light Absorption
Commercially available near-infrared lasers with linewidths less than 1 MHz provide
a straightforward means to address the D1 transition in K. In an atomic vapor, the
15
absorption of unpolarized photons near resonance as a function of propagation along
the the z-direction follows Beer’s Law
I(z) = I0e−naσ(ν)z (2.1)
where na is the alkali density, σ(ν) is the frequency dependant cross section, and I0
is the initial intensity. The total absorption over the entire propagation length L can
be interpreted in terms of the optical depth OD and is given by
OD = naσ(ν)L (2.2)
where OD 1 is referred to as “optically thick” and OD 1 is referred to as
“optically thin”. The frequency dependant cross section is given by
σ(ν) = crefL(ν) (2.3)
where c is the speed of light, re = 2.82 × 10−15 m is the classical electron radius,
and f is the oscillator strength of the transition2. The function L(ν) describes the
absorption lineshape
L(ν) =∆ν/2
(ν − ν0)2 + (∆ν/2)2. (2.4)
where ∆ν is the full-width at half maximum (FWHM) of the transition and ν0 is the
central frequency. On resonance, the cross section reduces to
σ(ν0) =2cref
∆ν(2.5)
2The oscillator strength of the D1 and D2 transitions are 1/3 and 2/3 respectively.
16
and integrating the cross section over all frequencies produces
σ =
∫ +∞
−∞σ(ν)dν = πcref. (2.6)
A derivation of these statements is available in Ref. [44].
Several mechanisms determine the width of the resonance including the natural
lifetime, pressure broadening, and doppler broadening. In the presence of several
amagats of 3He, ∆ν is dominated by the pressure broadening. The measured value3
is 13.2 GHz/amg for the FWHM [45]. The pressure of 3He is known at the time of
cell filling following the procedure described in Ref. [43].
At room temperature, K is a solid metal with a small vapor pressure (7×108/cm3).
Heating the cell to nearly 200C can significantly increase the vapor density to
1014/cm3. The density follows the well-known empirical expression
na =1026.2682−(4453 K)/T
(1 K−1)T/cm3 (2.7)
where T is the temperature in Kelvin [46]. In practice, the temperature of a vapor
cell indicates a 10-20% larger alkali density than is observed. The sensor is always
external to the cell and measures the temperature at a single location. It is believed
that the coldest region of the cell determines the alkali density. In addition, there
are a number of surface effects that allow alkali atoms to be adsorbed into the glass,
thereby lowering the effective density at a given temperature [47]. Nevertheless,
Eq. 2.7 provides an reasonable estimate of the density at a given temperature.
3This measurement is for the pressure broadening of K in a 4He environment. The value may bedifferent for 3He, but it is expected to be within a few percent.
17
2.1.3 Lightshifts
The presence of an oscillating light field on the atoms induces an ac-Stark shift on the
atomic vapor. This lightshift is equivalent to the effect of a dc magnetic field along
the direction of light propagation and is commonly referred to as a lightshift. The
energy shift is zero on resonance and dispersive in frequency
D(ν) =(ν − ν0)
(ν − ν0)2 + (∆ν/2)2(2.8)
rather than absorptive as in Eq. 2.4. The lightshift also depends on the degree of
circular polarization of light where linearly polarized light produces a zero lightshift.
In units of equivalent magnetic field, the lightshift can be expressed as [48]
~L =recfΦ~s
γeAD(ν) (2.9)
where Φ is the photon flux, A is the cross sectional area of the incident beam, γe is the
electron gyromagnetic ratio (Section 2.3), and ~s is the degree of circular polarization
of the light4. Considering the dispersive nature of the lightshift, it is useful to consider
the effects of both the closely spaced D1 and D2 transitions. They have opposite signs
and distinct oscillator strengths, so the net effect is
~L =recΦ~s
3γeA
(−D(νD1) + 2D(νD2)
)(2.10)
where the signs and oscillator strengths have been explicitly inserted. A semi-classical
derivation of this effect is available in Refs. [49, 50].
4Circularly polarized light provides |~s| = 1 and linearly polarized light is ~s = 0
18
mJ = -1/2 m
J =+1/2
4S1/2
4P1/2
σ+ P
umpin
g
Quen
chin
g
Quen
chin
g
Collisional Mixing
Spin Relaxation
Figure 2.2: Optical Pumping of K. Circularly polarized photons drive atoms from the 4S1/2
ground state to the 4P1/2 excited state. Collisions with noble gas atoms mix exited statepopulations. Nitrogen in the cell quenches relaxation from the excited state to the groundstate without reradiation of a photon. Atoms are always driven out of the mJ = −1/2ground state to eventually populate the mJ = +1/2 ground state. Image Credit: [1]
2.1.4 Optical Pumping
A complete description of optical pumping is available in Ref. [51]. For simplicity, we
ignore the nuclear spin and consider only pumping of the electron spin considering
only the 4S1/2 and 4P1/2 states in the simplified energy level diagram in Fig. 2.2. Un-
polarized atoms start with equal populations in the ground state sublevels. Circularly
polarized light on resonant with the 4S1/2 → 4P1/2 transition drives transitions with
selection rules ∆mJ = ±1. Left-hand, circularly polarized light σ+ drives transitions
from the mJ = −1/2 ground state to the mJ = +1/2 excited state. Collisions with
buffer gas atoms mix the excited state populations. 50 torr of N2 buffer gas provides
quenching without reradiation of a photon in a random direction. In this way, angu-
lar momentum is transferred from the photons to the vapor as evidenced by a large
fraction of the vapor pumped to the mJ = +1/2 ground state.
19
Optical pumping can be expressed quantitatively in terms of the polarization
d~P
dt= Rp(~sp − ~P )−Rsd
~P (2.11)
where |~P | = 1 corresponds to 100% polarization of the vapor, Rp is the pumping rate,
~sp is the degree of circular polarization, and Rsd is the spin-destruction rate. The
pumping rate is related to the intensity of the light as
Rp =Iσ(ν)
hν, (2.12)
and Rsd is a combination of several spin-destruction processes. For instance, collisions
with 3He, N2, and other K atoms transfer angular momentum to an unpolarized
spin of 3He to the rotational and translational degrees of freedom of the vapor in
the formation of a van der Waals molecule. For pumping along the the z-axis, the
equilibrium polarization becomes
Pz =Rp
Rp +Rsd
~sp · z. (2.13)
In principle, the pumping rate decreases as a function of the polarization of the vapor.
A circularly polarized pump beam propagating through an alkali vapor is described
by
dRp
dz= −naσ(ν)(1− Pz)Rp (2.14)
where only unpolarized atoms are pumped, and σ(ν) accounts for the detuning from
resonance. In the absence of optical pumping, Rp(z) is simply a decaying exponential
as in Eq. 2.1. In a fully polarized vapor, the pumping rate is constant. The general
20
solution to Eq. 2.14 can be written in terms of the product logarithm (ProductLog5)
Rp(z) = RsdProductLog[e−naσ(ν)z+Rp(0)/RsdRp(0)/Rsd
]. (2.15)
The typical polarization lifetime of a polarized K vapor is 70 ms. This is deter-
mined by
1
T1
= Rp +Rsd +Rwall (2.16)
where interaction with the wall of the vapor cell will also destroy the spin. The large
buffer gas pressures in the cell limit the diffusion of atoms to the walls. The diffusion
constant for K in a 3He buffer gas is
DK = 0.35 cm2/s
(√1 + T/(273.15 K)
nb/(1 amagat)
)(2.17)
where T is the temperature in Kelvin and nb is the density in amagats of 3He [52].
For a cell of several cm, this timescale is on the order of seconds, so this relaxation
can be safely ignored in this work. A thorough and instructive discussion of many of
these issues is available in Ref. [1].
A common method to monitor the degree of polarization of the vapor is through
optical rotation of linearly polarized light. The index of refraction for the left σ+
and right σ− circularly polarized light in a polarized medium is different depending
on the projection of the polarization along the direction of propagation. Linearly
polarized light is an equal superposition of both σ+ and σ− light, so the plane of
linear polarization will rotate through a polarized medium. In a polarized alkali-
metal vapor, a linearly polarized probe beam will rotate
φ =1
2lrecfnaPxD(ν) (2.18)
5The product logarithm is the inverse function of f(w) = wew. It is also known as the LambertW-function or omega function.
21
where l is the propagation length and Px is the projection of the polarization along
the propagation of the probe beam [48]. Notice that the optical rotation is zero if the
light is on resonance. This is because the real part of the index of refraction is equal
to 1 on resonance [44]. These ideas will be more fully explored in Section 2.4 in the
context of an atomic magnetometer.
2.2 Spin-Exchange Optical Pumping
When spin-polarized K atoms collide with the 3He buffer gas, spin angular momentum
transfers between the electron spin ~S in K and the nuclear spin ~K of 3He through
the hyperfine interaction Hhf = −A ~K · ~S in spin-exchange collisions. This is a well-
developed field and a comprehensive review is available in Ref. [53] and references
therein. Large 3He polarizations of up to 81% has been achieved in optimized systems
[54]. There are many details to consider, but this section serves to highlight aspects
of this technique applicable to the K-3He comagnetometer and the hyper-polarized
3He spin source.
2.2.1 Spin-Exchange Collisions
Figure 2.3 illustrates polarization of the 3He buffer gas where polarized K atoms collide
with unpolarized 3He. The total angular momentum is conserved in the collision and
spin polarization is transferred from the electron in K to the nucleus of 3He. The
closed electron shell in 3He isolates the nuclear spin from the environment such that
this process occurs over many hours. Similarly, once polarized, the 3He polarization
will relax over hours to days in a suitable environment. This is in stark contrast to
the typical polarization lifetimes of 70 ms in a polarized K vapor.
22
σ+
3He
K
K
K
K
3He
3He
K
K
3He
3He
a) b)
Bz
Figure 2.3: a) Circularly polarized photons polarize K spins through optical pumping. Col-lisions transfer angular momentum to the nucleus of 3He through the hyper-fine interaction.b) A spin-exchange collision preserves total angular momentum between K and 3He.
Given the long polarization lifetime of 3He, it is important to consider 3He diffusion
through the cell given by
D3He = 1.2 cm2/s
(1 + T/(273.15 K)
nb/(1 amagat)
)(2.19)
which is similar to Eq. 2.17 [52] and yields D3He ' 0.2 cm2/s for 10 amagats of 3He
at 180C. This implies that polarized 3He will explore the entire cell over its lifetime
over several hours. The 3He polarization ~P b is described by
d~P b
dt= Rne
se (〈~P a〉 − ~P b)−Rnsd~P b (2.20)
where this expression holds for a spin-1/2 nuclear spin6 and 〈~P a〉 represents a spatial
average of the K polarization. The nuclear spin-exchange rate Rnese can be determined
in terms of an effective cross section
Rnese = naσsev (2.21)
6An extra factor is required for nuclei with higher moments [53].
23
where7 v =√
8kT/(πm) and m is the reduced mass of K and 3He. The nuclear
spin-destruction rate Rnsd is a combination of a spin-destruction cross section with K
where the the angular momentum is lost to the vapor and some specific 3He relaxation
mechanisms described in the next section.
2.2.2 Longitudinal Relaxation: T1
Long T1 times exceeding several days have been observed in hyper-polarized 3He
systems. The dominant contributions to this relaxation is from 3He dipole-dipole
collisions, relaxation along magnetic field gradients, and wall relaxation. The total
relaxation rate observed is a contribution of each of these mechanisms
1
T1
=1
T dd1
+1
T∇1+
1
Twall1
. (2.22)
The dominant relaxation in the spin source is the dipole-dipole relaxation given by
1
T dd1
=nb
744 amagat · hour(2.23)
where nb is the density in amagats [55]. The 12 amagat spin source exhibits relax-
ation times near the 60 hour limit determined by the dipole-dipole collision rate.
In the K-3He comagnetometer, magnetic field gradients are the dominant relaxation
mechanism. This rate is given by
1
T∇1= D3He
|∇B⊥|2B2z
(2.24)
where the notation |∇B⊥|2 = |∇Bx|2+|∇By|2 and Bz represents the magnitude of the
magnetic field along the direction of polarization [56, 57]. Diffusion along transverse
magnetic field gradients leads to relaxation of the spins. The gradients in the case of
7The notation for Rnese will clearer in the context two spin ensembles in Section 2.5.2 and thedescription of the K-3He comagnetometer in Chapter 3. For instance, Rense = nbσsev.
24
the comagnetometer are limited by the nonuniform dipolar field created by uniformly
polarized 3He with more details provided in Section 4.6.1.
Despite the isolation of the 3He nucleus by the closed shell of electrons, collisions
with the wall can lead to further relaxation. This process is little understood, but the
most comprehensive and recent 3He wall relaxation studies appear to be in Refs. [58,
59, 60] and references therein. There is some “black magic” in cell manufacture to
avoid issues of wall-relaxation. In most cases in this work, significant wall relaxation
has been avoided except as described in Section 4.6.2.
Including all of these effects, the equilibrium 3He polarization with a polarized K
vapor along the z-direction is given simply by
P bz =
Rnese
Rnese +Rn
sd
〈P az 〉 (2.25)
and reaches equilibrium in several time constants of 1/(Rnese +Rn
sd).
2.3 Larmor Precession
The Lorentz violation search attempts to measure an energy shift in the spin preces-
sion frequency as an experimental signature of new physics. Both the K and 3He spins
have an associated magnetic moment coupling to a magnetic field as does a classical
magnetic dipole. Under the influence of a magnetic field, the spin will precess at
the Larmor frequency. The following sections provide details on this phenomena and
clearly define the sign convention of spin precession followed in this work.
2.3.1 Isolated Spin
The interaction Hamiltonian for a magnetic moment in a magnetic field is
H = −~µ · ~B. (2.26)
25
y
z
x
<S>
ω
B
Figure 2.4: Evolution of 〈~S〉 in a Bz magnetic field as in Eq. 2.28. Here, the magneticmoment is anti-aligned with the spin.
For the case of the free electron, ~µ = −gSµB ~S where gS ≈ 2 and µB = 9.28476 ×
10−24J/T is the Bohr magneton and yields
H = gSµB ~S · ~B. (2.27)
The time evolution of this spin in a magnetic field is known as Larmor precession.
Using the time-dependant Schrodinger equation, the following equation of motion
describes the expectation value of a spin in a magnetic field
d〈~S〉dt
=gSµB~
~B × 〈~S〉 = γe ~B × 〈~S〉 (2.28)
where γe is the electron gyromagnetic ratio. This agrees with the classical result of
the torque experienced by a magnetic moment in a magnetic field. The electron will
precess about a magnetic field at frequency ω = γe| ~B| (Fig. 2.4).
Rotating Frame
Classically, the time derivative of any time-dependent vector in the lab frame d ~A/dt
and the time derivative of the same vector in the rotating frame ∂ ~A/∂t are related
26
by
d ~A
dt=∂ ~A
∂t+ ~Ω× ~A (2.29)
where ~Ω is the rotation vector. In the rotating frame, Eq. 2.28 transforms to
∂〈~S〉∂t
= γe
(~B −
~Ω
γe
)× 〈~S〉. (2.30)
Throughout this thesis, we use total derivatives to denote physics in the lab frame
and partial derivatives to denote physics in the rotating frame. There are many roads
to this result, both classical and quantum mechanical with a good summary provided
in Ref. [61].
2.3.2 Bound Electron
In atomic K, the total angular momentum of the electron modifies the precession fre-
quency depending on the atomic state and can be determined following the argument
from Ref. [62]. The simplest case considers the electron magnetic moment as the sum
of orbital and spin angular momenta
~µ = −µB(gS ~S + gL~L) = −µB( ~J + ~S) (2.31)
where as usual gS ≈ 2 and gL = 1. Here, ~J = ~L+ ~S is a “good” quantum number, so
~S must be evaluated in the J-basis. It follows that
〈~µ〉 = −µB(〈 ~J〉+ 〈~S〉
)= −µB
(1 +
〈~S · ~J〉J(J + 1)
)〈 ~J〉. (2.32)
27
A specific case of the well-known Wigner-Eckhart Theorem known as the Projection
Theorem [14] gives
〈α′, jm′|Vq|α, jm〉 =〈α′, jm| ~J · ~V |α, jm〉
j(j + 1)~2〈jm′|Jq|jm〉, (2.33)
to evaluate 〈~S〉 in the J-basis. The product 〈~S · ~J〉 is straightforward to evaluate to
yield the g-factor for this state ~µ = −gJµB ~J where
gJ = 1 +J(J + 1) + S(S + 1)− L(L+ 1)
2J(J + 1). (2.34)
This result modifies Eq. 2.28 by replacing gS with gJ in the gyromagnetic ratio and
modifies the precession depending on the electronic state.
For K in the 4S1/2 ground state, J = 1/2, L = 0, and S = 1/2 which yields gJ = 2,
the same result as the g-factor for the free electron.
Nuclear Spin Contribution
The argument in the previous section neglects the contribution to the total angular
momentum from the nucleus. For a complete treatment, the hyperfine interaction
must be included in the magnetic moment
~µ = −gFµB ~F = −µB(gJ ~J + gI~I) (2.35)
where ~F = ~J + ~I. Since the nuclear moment is several orders of magnitude smaller
than the Bohr magneton, the approximation gI/gJ 1 neglects the second term in
Eq. 2.35. Similar to Eq. 2.32, ~J must be expressed in terms of ~F . Once again, the
Projection Theorem proves useful with
〈 ~J〉 =〈 ~J · ~F 〉F (F + 1)
〈~F 〉 (2.36)
28
where 〈 ~J · ~F 〉 is simple to evaluate in the ~F -basis. This produces
gF = gJ =(F (F + 1) + J(J + 1)− I(I + 1)
2F (F + 1)
)(2.37)
for easy replacement in Eq. 2.28 for any well-defined total angular momentum state.
Considering the general case of the outer shell electron in an alkali atom in the
S1/2 ground state where J = 1/2 and F = I ± 1/2, produces the well-known result
∂〈~S〉∂t
=gFµB~
~B × 〈~S〉 = ± γe2I + 1
~B × 〈~S〉. (2.38)
The rate of electronic precession appears to “slow down” in response to dragging the
heavy nuclear spin. The slowing down increases for larger nuclear spins. The valence
electron precesses in opposite directions depending on the hyperfine state, and the
state with maximal angular momentum F = I+1/2 follows the same sign convention
as the free electron. For 39K, I = 3/2, so γeff = γe/4.
The precession of the electron in K is modified by its nonzero nuclear spin and
the direction of precession depends on the hyperfine level. The spin precession of 3He
is much less complex. In the ground state, the outer shell elections are paired so only
the spin-1/2 nucleus is free to precess. The nuclear magnetic moment is a factor of
103 smaller than the electron magnetic moment and will precess at a lower frequency.
The gyromagnetic ratios take the values
γe =gµB~
= 2π × 2.8 MHz/gauss (2.39)
γn =|µ3He|~
= 2π × 3.243 kHz/gauss (2.40)
where we have selected the convention that the gyromagnetic ratio is a positive num-
ber. This explicit definition will help with clarity in the equations that follow. Both
29
the electron in K and nucleus in 3He have negative magnetic moments, so the preces-
sion in a magnetic field follows the sign of Eq. 2.28.
2.4 Alkali-Metal SERF Magnetometer
Atomic alkali-metal magnetometers use the fundamental precession of the outer shell
electron spin in a polarized alkali-metal vapor to determine the local magnetic field.
This idea was first proposed by Dehmelt [63] and achieved by Bell and Bloom [64].
The atomic magnetometer has since become a well-developed technology.
Over the decades, systematic improvements in this simple idea have increased
sensitivity and reduced technical noise. Most recently, the development of the spin-
exchange relaxation free (SERF) magnetometer has extended sensitivity below the
fT level [65]. This magnetometer circumvents spin-exchange relaxation to be limited
by the slower spin-destruction relaxation. It has achieved some of the best magnetic
field measurements to date and has reached the level of 0.16 fT/√
Hz in a gradiometer
arrangement [66]. The comagnetometer described in this thesis inherits its sensitivity
from the SERF, so it suffices to discuss the relevant details here.
2.4.1 Magnetometry
The classic magnetometer scheme is to optically pump an alkali metal vapor with
circularly polarized light on resonance (Fig. 2.5). The spins aligned along the z-
direction precess under an applied magnetic field in the y-direction (Section 2.3). A
linearly polarized off-resonant probe beam measures the projection of the spin along
the x-direction (Eq. 2.18). A simple fixed polarizer translates the optical rotation
into intensity of the light which is easily measured on a photodiode. This is referred
to as the magnetometer signal.
30
By
S
Probe
Pump
σ+
y
z
x
Figure 2.5: A circularly polarized pump beam polarizes the atoms along the z-direction. Amagnetic field in the y-direction causes the spins to precess into the x-direction where thex-projection of the spin is measured through optical rotation of a linearly polarized probebeam.
Ignoring factors of 2 and the magnitude of the spin, the fundamental sensitivity
of an atomic magnetometer is given by
δB ' 1
γ√NT2t
(2.41)
where γ is the Larmor frequency of the atoms, N is the number of atoms in the
measurement volume, and T2 is the coherence time for the precession and the spins.
The sensitivity improves as the square root of the measurement time t. For many
years it was believed that T2 was always limited by spin-exchange collisions 1/T2 ∼ Reese
constraining the best sensitivity to 1 fT cm3/2/√
Hz.
2.4.2 SERF Regime
The 4S1/2 ground state in 39K splits into two hyperfine states F = 1 and F = 2. As
shown earlier in Section 2.4, each of these states precess in the opposite direction.
We can think of this precession as an effective vector (Fig. 2.6a). In an atomic vapor,
these atoms undergo spin-exchange collisions in which the total angular momentum
mF1 + mF2 is conserved. This implies that an individual atom has a probability
31
mF=2
mF=1
B
+ω
-ω
(a)m
F=2
mF=1
B
+ωQ
(b)
-2-1
-10
+1
0+1
+2F=2
F=1
(c)
-ω
+ω
4S1/2
Pe
˜50%
ω Rse
Spin-exchangerelaxation
Spin-exchange
Figure 2.6: a) Precession of the F = 1 and F = 2 states in the 4S1/2 ground state interms of an effective vector. Spin-exchange causes atoms to switch hyperfine levels andprecess in opposite directions. b) Precession of the hyperfine levels in the SERF regime.Spin-exchange occurs more rapidly than precession. Due to the the higher statistical weightin the F = 2 state, the net precession is in this direction. c) Population of the hyperfinesublevels at P e ∼ 50%.
of switching hyperfine levels and thus switches spin precession direction with every
spin-exchange collision. The switching of precession direction leads to dephasing in
the spin precession of the ensemble to limit T2 and the magnetometer sensitivity
(Eq. 2.41).
It was observed experimentally [67] and later explored theoretically [68] that a
high-density, optically-pumped atomic vapor in a low magnetic field exhibits a longer
T2 than expected from consideration of the spin-exchange rate. In the limit where
Reese ω0, the atoms switch hyperfine states faster than they can precess in the
magnetic field. Under optical pumping conditions, the atoms are pumped towards
the |F = 2,mF = 2〉 end state which carries a greater statistical weight. In this way,
the atoms lock to a net precession at a slower rate in the direction of the F = 2 state
depending on the polarization (Fig. 2.6b).
A precise explanation for the precession of the atoms in the SERF regime is
described in [69]. To determine the net precession at an arbitrary polarization, a
spin-temperature distribution is assumed (Fig. 2.6c). For a spin-3/2 nucleus, we
32
arrive at the following expression
ω/ωeff = 2− 4
3 + P 2(2.42)
where the ωeff = γeff | ~B|. This effect serves to further “slow down” the precession of
the outer electron in an alkali atom. We define the following “slowing-down factor”
for convenience
Q(P ) = 4/(2− 4
3 + (P )2
)(2.43)
where the leading 4 comes from the argument in Section 2.3.2 which is further modified
by the SERF regime. Q(P ) varies between 4 for full polarization and 6 for zero
polarization. Alkali-metal magnetometers achieve maximum sensitivity at P = 50%
where Q(0.5) = 5.2. This factor slows down all of the rates for the alkali atom in the
SERF regime.
Despite the rapid switching of hyperfine states, it is interesting that the precession
frequency is still on the same order of precession in the absence of spin-exchange col-
lisions. However, this minor change in the precession frequency produces a dramatic
change in the coherent precession of the spins.
The spin-exchange rate that contributes to the total relaxation of the spins be-
comes
Reese =
(γeB
Q(P )
)2[Q(P )]2 − (2I + 1)2
2Rse/Q(P )(2.44)
in the limit of low magnetic field [70]. The distinguishing feature of this regime is the
dependance on the square of the magnetic field. The increase in T2 improves the atom
shot noise limit for magnetic sensitivity (Eq. 2.41). Spin-destruction collisions Reesd
have a cross section several orders of magnitude smaller than spin-exchange collisions
and limit the fundamental sensitivity to 0.02 fT cm3/2/√
Hz.
33
2.4.3 SERF Magnetometer
The SERF magnetometer is a vector magnetometer meaning that it is most sensitive
to a particular direction of the magnetic field. In the SERF regime, spin precession
is slow, so the spins precess in a small applied field by an angle
φ ∝ T2γeBy (2.45)
in the steady state8. A 10 fT field and coherence time of 3 ms gives a spin precession
angle of 5 × 10−6 rad. Detection of optical rotation of linearly polarized light near
the atomic resonance provides a means to measure such small angles (Eq. 2.18).
The behavior of a magnetometer can be well-described by the phenomenological
Bloch equation describing the precession of a spin-1/2 particle in a magnetic field
∂ ~P
∂t=
1
Q(P )
[γe ~B × ~P +Rp
(z − ~P
)−Rsd
~P]
(2.46)
where Rp is the pumping rate, Rsd is the spin-destruction rate, and Q(P ) is defined
in Eq. 2.43. Through optical pumping, the equilibrium polarization reaches
Pz =Rp
Rp +Rsd
=Rp
Rtot
. (2.47)
The signal comes from the projection of the spins along the x-axis. The steady state
polarization in this direction is
Px = PzγeByRtot + γeBxBz
R2tot + γe
(B2x +B2
y +B2z
)/Rtot
(2.48)
8There is no factor of Q here since it modifies both γe and T2.
34
where we have expressed the rates in terms of Rtot = Rp +Rsd. In the limit that the
magnetic fields are small, Bx, By, and Bz γe/Rtot, the leading behavior is
Px =PzγeRtot
By (2.49)
demonstrating the primary sensitivity of the SERF to a By field. The coefficient
Pzγe/Rtot determines the sensitivity to magnetic fields. For fixed Rsd, the maximum
sensitivity is reached when Rp = Rsd at Pz = 50%. The high sensitivity of the SERF
comes from the dramatic reduction of Reese in Rtot from Eq. 2.44.
2.5 Interacting Spin Ensembles
Spin-exchange optical pumping (Section 2.2) transfers angular momentum from po-
larized K to the nuclear spins in 3He. These co-located spin ensembles are coupled
through spin-exchange and dipolar magnetic interactions. The spatial overlap en-
hances the magnetic interaction between spins compared to the case of an isolated
spin. We observe the interaction of the spins by treating the K as a magnetometer
(Section 2.4) and measuring the evolution of P ex through optical rotation of a linearly
polarized probe beam. The interactions described here will form the basis for the
K-3He comagnetometer described in Chapter 3.
2.5.1 Contact Interaction
The magnetization of a polarized atomic species generates a dipolar magnetic field.
It is well known that the magnetic field from a uniformly magnetized sphere (in cgs
units) is given by
~B =8π
3~M (2.50)
35
where ~M is the magnetization density vector9. In the case of hyper-polarized K and
3He, the magnetic field of one species causes a shift in the precession frequency of the
other.
Collisions within the vapor modify the magnetic interaction. Spatial overlap of the
valence electron in K with the nucleus of 3He produces an enhancement in the effective
magnetic field experienced by both the K and 3He. The Fermi-contact hyperfine
interaction α(R) ~K · ~S mediates the interaction between the noble-gas nuclear spin ~K
and alkali-metal electron spin ~S
α(R) =8π
3gsµB
µKK|ψ(R)|2 (2.51)
where |ψ(R)|2 is the square of the wavefunction for the valence electron in K at the
3He nucleus [71]. This interaction is well-described by effective magnetic fields with
the appropriate enhancement factors [71],
δ ~Ba =8π
3
µKKκabnb〈 ~K〉 (2.52)
δ ~Bb = −8π
3gSµBκbana〈~S〉 (2.53)
where δ ~Ba is the shift for the alkali-metal spins and δ ~Bb is the shift for the noble-
gas spins. The enhancement factors κab and κba are defined as the ratio of field
experienced by the alkali-metal/noble-gas spins to the macroscopic field produced
by a spherical cell containing the magnetization of a noble-gas/alkali-metal at the
corresponding density and polarization [72].
The enhancement factors κab and κba contain contributions from alkali-metal–
noble-gas binary collisions κ0 and van der Waals molecules κ1. Noble-gas pressures
above an atmosphere decrease the lifetime of van der Waals molecules such that both
9The analogous expression in SI units is ~B = 23µ0
~M where µ0 is the permeability of free space.
36
κab and κba converge to a single value of κ0 [71]. This is the relevant regime of this
work.
The interaction potential between atoms in collisions leads to a slight temperature
dependence. This is well-measured over the range of temperatures used in spin-
exchange optical pumping [73] and is given by
κ0 = 5.99 + 0.0086[T − 200(C)]/C. (2.54)
For notational convenience, we use a simple expression to describe the field experi-
enced by the electron and nuclear spins ~Be/n. For the case of a uniformly magnetized
sphere,
~Be =λM e ~P e (2.55)
~Bn =λMn ~P n (2.56)
where λ = 8πκ0/3 is a geometrical constant containing the enhancement factor,
M e = −gSµBna and Mn = µKnb/K = µ3Henb give the total magnetization density
of the cell, and ~P e/n is the electron/nuclear polarization vector. For reference, the
relevant magneton values are
µB = 9.274× 10−24 J/T (2.57)
µ3He =− 1.074× 10−26 J/T. (2.58)
Note that both the magnetic moments of optically pumped K and 3He are negative;
this implies that M e and Mn are negative.
37
2.5.2 Coupled K-3He Dynamics
The dynamics of overlapping K and 3He spin ensembles can be modeled with a sim-
plified set of coupled Bloch equations [70]. The interaction between spins enters as
the spin of one species precessing in the magnetization of the other (Section 2.5.1)
and monitored through P ex as in the SERF (Section 2.4). Following the spin evolution
conventions in Section 2.3, the simplified Bloch equations are
∂ ~P e
∂t=γeQ
( ~B + λMn ~P n)× ~P e +Rtot(Pe0 z − ~P e)/Q (2.59)
∂ ~P n
∂t= γn( ~B + λM e ~P e)× ~P n +Rn
tot(Pn0 z − ~P n) (2.60)
where the spins relax to the equilibrium polarizations of P e0 and P n
0 for the electron
and nuclear spins respectively. The relaxation rates Rtot and Rntot refer to general-
ized relaxation rates for the electron and nuclear spins. Also, the magnitude of the
alkali polarization is assumed to remain constant such that the slowing down factor
(Eq. 2.43) decouples from the polarization Q(P e) → Q. Here, we observe that the
spins precess in different fields; the electron precesses in the sum of the applied field
and magnetization of the nuclear spins while the nuclear spins precess in the sum of
the applied field and the magnetization of the electron spins. The electron and nu-
clear spins experience the magnetic field from their own magnetization, but this effect
is not listed in Eqs. 2.59 and 2.60. This field produces no torque (~P e/n × ~P e/n = 0)
and is omitted.
To study small transverse excitations of the spins, these equations can be lin-
earized. Shifting to complex form P e⊥ = P e
x + iP ey and P n
⊥ = P nx + iP n
y with the
assumption that |P e⊥| P e
0 and |P n⊥| P n
0 yields the linear coupled differential
38
equations
P e′⊥ =
(iγe(B
az + λMnP n
z )P e⊥ −RtotP
e⊥ − iγe(λMn
z Pn⊥)P e
z
)/Q (2.61)
P n′⊥ = iγn(Ba
z + λM eP ez )P n
⊥ −RntotP
n⊥ − iγn(λM e
zPe⊥)P n
z (2.62)
where Baz is an applied magnetic field along the z-direction. Linear algebra techniques
efficiently solve for P e⊥ and P n
⊥.
The exact solution takes the following form in terms of the eigenmodes of the
system
P ex(t) = Re[P1e
(Ae+An−F )t/2 + P2e(Ae+An+F )t/2], where (2.63)
Ae =(iγe(B
az + λMnP n
z )−Rtot
)/Q, (2.64)
An = iγn(Baz + λM eP e
z
)−Rn
tot, and (2.65)
F =√
(Ae − An)2 − 4(γeλM eP ez )(γnλMnP n
z )/Q. (2.66)
Two distinct oscillation frequencies persist (Ae+An+F )/2 and (Ae+An−F )/2 cor-
responding to contributions from both nuclear and electron spins. For large magnetic
fields, where Baz is much larger than the magnetization of either species, the response
is the sum of both spins precessing independently in Baz . The nuclear and electron
spins are only loosely coupled through F and relax at their respective rates.
When the alkali atoms experience a small magnetic field, they enter the SERF
regime (Section 2.4). We adjust Baz to the following condition
Bac = −λM eP e
z − λMnP nz (2.67)
to reduce the magnetic field experienced by the alkali from the magnetization of the
3He. We refer to Bac as the compensation point. Here, the spins precess in an effective
39
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Px (
Arb
. sc
ale)
Time (s)
B
a=Bc+80 μgauss
Ba=Bcz
za
a
e
Figure 2.7: Decay of the spins from an initial small tipping angle. At the compensationpoint (solid), the damping time is dramatically shortened over a small detuning (dashed).
field equal to their own magnetization
Ae → iγe(−λM eP ez )/Q−Rtot/Q (2.68)
An → iγn(−λMnP nz )−Rn
tot (2.69)
and under typical comagnetometer conditions, these frequencies are on the order of
tens of Hz and are nearly equal
γnλMnP n
z ≈ γeλMeP e
z /Q. (2.70)
In general, electron and nuclear precession frequencies are distinct by several orders
of magnitude. Here, the spins experience different fields and precess together to form
a hybrid resonance. The strong coupling in this resonance leads to some dramatic
behavior in the evolution of the spins.
Analytic expressions for the oscillatory and damping response of the comagne-
tometer can be determined by separating the exponents in Eq. 2.63 into real and
imaginary parts
iω + Γ = (Ae + An ± F )/2. (2.71)
40
Due to the square root in F , simple analytic expressions at the compensation point can
be determined with the reasonable approximations γeλMeP e
z /Rtot 1 and Rntot
Rtot. The most dramatic behavior occurs for the decay of the the nuclear spins,
Γn = −(γeλMeP e
z /Q)(γnλMnP n
z )Rtot/Q
R2tot/Q
2 + (γnλMnP nz )2
(2.72)
where the damping rate is no longer determined by Rntot, but by the product of
the magnetizations and the alkali relaxation rate. Under typical comagnetometer
conditions, this is on the order of several seconds instead of the more usual hundreds
of seconds. The nuclear spins precess in a field of the same magnitude as their
magnetization
ωn = −γnλMnP nz
(1 +
(γeλMeP e
z /Q)(γnλMnP n
z )
R2tot/Q
2 + (γnλMnP nz )2
)(2.73)
with a correction for the presence of the alkali magnetization. The alkali damping
rate continues to be dominated by Rtot/Q
Γe = −Rtot/Q+(γeλM
eP ez /Q)(γnλM
nP nz )Rtot/Q
R2tot/Q
2 + (γnλMnP nz )2
(2.74)
with a small reduction similar to Eq. 2.73. The precession of the spins
ωe = − (γeλMeP e
z /Q)R2tot/Q
2
R2tot/Q
2 + (γnλMnP nz )2
(2.75)
is mostly unaffected by the presence of the nuclear spin species. The slowest of the two
frequencies is typically ωn on the order of 10 Hz. We refer to ωn as the compensation
frequency.
41
2.5.3 Oscillatory Magnetic Field Suppression
It is interesting to consider the behavior of the coupled spins at the compensation
point under the influence of a time-dependant magnetic field. Inserting an oscillatory
field B⊥(t) =(Bx + iBy
)cos (ωt) into Eqs. 2.61 and 2.62 yields the following set of
linearized equations
P e′⊥ =
(iγe(B
az + λMnP n
z )P e⊥ −RtotP
e⊥ − iγe
(λMn
z Pn⊥ +B⊥(t)
)P ez
)/Q (2.76)
P n′⊥ = iγn
(Baz + λM eP e
z
)P n⊥ −Rn
totPn⊥ − iγn
(λM e
zPe⊥ +B⊥(t)
)P nz (2.77)
which can be solved in a straightforward manner for oscillatory solutions in complex
form. It is important to consider both co-rotating eiωt and counter-rotating e−iωt
solutions. Enforcing the compensation condition (Eq. 2.67) yields
P e⊥(t) = −1
2
[P ez γeω(Bx + iBy)e
iωt/Q
(γnλMnP nz + ω)(ω − iRtot/Q) + ωγeλM eP e
z /Q
+P ez γeω(Bx + iBy)e
−iωt/Q
(γnλMnP nz − ω)(ω + iRtot/Q) + ωγeλM eP e
z /Q
](2.78)
where it has been assumed that Rntot ' 0.25 hrs−1 is much less than all the other
rates10. The two terms in Eq. 2.78 are equivalent for the replacement ω → −ω as
expected for the co-rotating and counter-rotating solutions.
It will be common for the terms λMnP nz and λM eP e
z to appear repeatedly. Solu-
tions become considerably simpler to interpret with the replacement
Be =λM eP ez (2.79)
Bn =λMnP nz (2.80)
10Equation 2.78 corrects the statement in Eq. 2.119 of Ref. [28] by including both the two rotatingsolutions and the inclusion of Q.
42
where bothBe andBn are negative following the convention described in Section 2.5.1.
In the limit that ω |γnBn|
P ex(t) ≈ P e
z γeRtot
(ωBx sin(ωt)
γnBn+O(ω2)
)(2.81)
for a slowly modulated Bx field where the 3He suppresses the effects of the applied
magnetic field at low frequencies. The same is true for a slowly modulated By field
P ex(t) ≈ −P
ez γeRtot
(ω2By cos (ωt)
(γnBn)2+O(ω3)
)(2.82)
where the low frequency behavior is further suppressed by an additional factor of
(ω/γnBn). We expect that a slowly modulated Bz field will be further suppressed
because it cannot provide a torque on spins oriented along the z-direction to precess
into the x-direction where they can be detected.
The fast damping of nuclear spins and suppression of slowly varying magnetic
fields are fundamental features of these coupled spin ensembles. In the following
chapter, we explore the consequences of their combined effect to describe a K-3He
comagnetometer.
43
Chapter 3
Next Generation K-3He
Comagnetometer
A comagnetometer is combination of two magnetometers using two different spin
species. A magnetic field will interact with each species in a well-defined way such
that a difference or a ratio of each response separates magnetic interactions from other
effects. The K-3He comagnetometer described in this chapter provides suppression
of magnetic fields, but remains sensitive to non-magnetic spin couplings. It is well-
suited for Lorentz violation or long-range spin-dependant force searches described in
Chapters 5 and 6.
3.1 Comagnetometer Overview
The K-3He comagnetometer contains two overlapping spin ensembles of K and 3He
in a spherical glass vapor cell. We typically heat the cell to 180C to produce a
high-density alkali vapor of 7 × 1013/cm3 from several mg of K metal in the cell.
Optical pumping polarizes the electron spins in a K vapor. Spin-exchange collisions
with several amagats of 3He buffer gas (1× 1020/cm3) hyper-polarize several percent
of these nuclear spins. An off-resonant linearly polarized probe beam orthogonal to
44
the optical pumping direction continuously measures the projection of the K spins
along the probe propagation axis through optical rotation of the linearly polarized
light (Eq. 2.18).
This alkali-metal–noble-gas pair has been chosen for the small spin destruction
cross section of K. These spins form a joint resonance at the compensation point (Sec-
tion 2.5.2) where the comagnetometer also inherits the high sensitivity of the SERF
magnetometer (Section 2.4.3). The coupled dynamics strongly damps excitations of
the nuclear spins within a few seconds to quickly bring the system to the steady state.
The response to slowly varying magnetic fields decreases for decreasing frequency as
the 3He buffer gas suppresses the effects of a magnetic field (Section 2.5.3).
This technique is unique in that it uses a pair of co-located electron and nuclear
spins as a comagnetometer. Several experiments have used pairs of co-located nuclear
spins as a comagnetometer [29, 33, 40]. A pair of electron and nuclear spins has been
used previously [34], but these spins were located in separate vapor cells. No others
systems have yet combined electron and nuclear spins in the same vapor cell as with
the K-3He comagnetometer. The response of the K-3He comagnetometer will be
described through a picture (Section 3.1.1), then through a more technical treatment
of the Bloch equations (Section 3.1.3).
3.1.1 Comagnetometer Description
Figure 3.1a illustrates the configuration of the K-3He comagnetometer. K atoms are
pumped along the z-direction and probed along the x-direction. At the compensation
point, the applied magnetic field Bac cancels the magnetization of the electron and
nuclear spins to place the K in the SERF regime. In the case of a slowly ω ωn
applied Bx Bac perturbation, the K spins are continuously repolarized and have
no long term memory while the 3He atoms observe the changing field and precess in
response (Fig. 3.1b). The 3He spins quickly damp to align with the total applied field.
45
Since 3He has its spin oppositely oriented to its magnetic moment, Bn cancels the
Bx perturbation to leading order, so the K atoms experience no change in magnetic
field. The same holds for a small By perturbation; however, a small change in Bz is
uncompensated. A small change in Bz provides no torque on the spins, so no signal
can be observed. An initial misalignment of the spins in the x- or y-direction will lead
to sensitivity in Bz, but only as a product with the small misalignment (ie. ByBz).
The comagnetometer suppresses the effects of magnetic fields to leading order, but
imperfect cancelation can lead to such second order effects.
For the comagnetometer to be useful for anomalous spin searches such as for a
Lorentz-violating field, it must preserve sensitivity to the anomalous field. Consider
the case of a field that couples only to electron spins ~βe and applied along the y-
direction (Fig. 3.1c). The noble-gas does not precess to follow this field because there
is no coupling. The K spins precess to project onto the probe direction. This gives
a signal response proportional to the field. Now, consider an anomalous field that
couples only to the nuclear spins ~βn applied along the y-direction (Fig. 3.1d). The
noble-gas spins follow the net projection of both magnetic and anomalous fields. From
the point of view of the K atoms, the magnetization of 3He is uncompensated and
provides a real magnetic field causing the K spins to precess. The magnetization of
3He is opposite to its spin, so the K responds in the opposite direction under ~βn as it
does under ~βe.
We expect the leading order behavior of the comagnetometer be βey − βny with no
first order magnetic field sensitivity. One way to think about this is to consider the
case where the anomalous field couples to the spins in the same way as a magnetic
field, that is, in proportion to the magnetic moment of the spin. In this case, βey = βny ,
so there is no leading order sensitivity. However, if the coupling differs from that of
a magnetic field, then the K spins will precess in response to either the electron or
46
BxBc
B
K
Pump
Pro
be
(b)
Bc
K
Pump
Pro
be
βe(c)
Bc
Bn
Bc
K
Pump
Pro
be
(a)
βn
B
PumpPro
be
(d)
K
Bc
Bn B
n
Bn3He
3He
3He
3He
x
zy
a
aa
a
a
Figure 3.1: a) K-3He comagnetometer at the compensation point. Bac compensates for Bn
such that the K atoms see the low field condition of SERF. b) A small adiabatically appliedBx field rotates the total applied field vector. The 3He rotates to follow the total field and itseffective magnetic field cancels the change in magnetic field seen by the K. c) An anomalousfield perpendicular to the pump and probe beams couples only to the K and rotates thespins into the probe beam. This leads to a signal response. d) An anomalous field couplesonly to the nuclear spins. The spins follow the combined magnetic and anomalous field androtate out of the page. This leaves an uncompensated magnetic field projecting into thepage. The K spins rotate under this field and project onto the probe beam in the oppositedirection as in image (c).
47
nuclear spin coupling. In this way, the balance between the first order suppression of
magnetic fields is broken.
In practice, the comagnetometer provides 103− 104 suppression of magnetic fields
while remaining sensitive to the anomalous fields of interest. Further suppression
of magnetic interactions can be introduced by enclosing the comagnetometer within
many layers of magnetic shielding. The inherent magnetic field suppression of the
comagnetometer provides shielding from thermal Johnson noise symptomatic to many
magnetic shield materials.
This description of the comagnetometer relies on pictorial representation of the
spin and associated magnetization vectors. A calculation of the imperfect compen-
sation is provided in Appendix C. This will confirm the validity of this picture with
the full solution in Section 3.2. The intuition gained through the picture provides a
better understanding and interpretation of the full analytical solution.
3.1.2 Practical Anomalous Fields
We define an anomalous field as any that couples to the spin differently than in
proportion to the magnetic moment. Such an interaction for the electron yields the
Hamiltonian
H =gsµB~
~B · ~S + η~βe · ~S (3.1)
where η casts the anomalous field ~βe into units and sign of an effective magnetic
field. Lorentz-violating or other spin-dependant forces have been proposed, but not
observed (Section 1.2 and 6.5.2). There are several well-known anomalous fields
relevant to comagnetometer operation, namely lightshifts and rotations.
A lightshift comes from the ac Stark shift of the oscillating light field (Sec-
tion 2.1.3). This effect shifts the energy levels much the same way as a magnetic
field in the vicinity of an electronic transition. The on-resonant pump and near-
48
Bc
Bn
Bc
K
Pump
Pro
be
(a)
3He
x
z
y
Ω
Bc
Bn
Bc
K
Pump
Pro
be
(b)
3He
x
z
y
Ω
Bx Bx
a
a
a
a
Figure 3.2: a) Rotation of the comagnetometer with respect to an inertial frame. Theelectron spins are quickly repolarized while the nuclear spins remain oriented with respectto an inertial frame. A small Bx component of the compensation field deviates from thedirection of the spins. b) The 3He responds to the Bx field by precessing into the page. Themagnetization extends out of the page and causes the K spins to project onto the probebeam. A positive rotation responds with the opposite sign as βny (Fig. 3.1d).
resonant probe beams introduce such lightshifts to the K spins only. These fields
are too far off-resonant to significantly influence nuclear spin transitions in 3He. The
lightshift ~L presents itself as an anomalous field coupling only to electrons just as in
Fig. 3.1c.
Furthermore, a rotation about an inertial frame at a rate slower than the compen-
sation frequency introduces an anomalous field. Spins carry angular momentum and
will maintain the same direction with respect to an inertial frame under a rotation
of the comagnetometer. The case where the comagnetometer is rotated about the
y-direction is shown in Fig. 3.2a. The K spins are quickly repolarized along the new
direction by the pump laser; however, the 3He spins remain in the inertial frame. The
3He experiences a torque from the reoriented compensation field and precesses into
the page (Fig. 3.2b). The magnetization projects out of the page and causes the K
atoms to precess into the probe beam. The response of a positive rotation around the
y-direction results in the opposite response as the the response from an anomalous
49
nuclear field in Fig. 3.1d and defines an equivalent anomalous field due to rotations as
−~Ω/γn. In fact, precession of both spin species contributes to the rotation response;
however, the response is dominated by the nuclear spins because γe/Q(P e) γn.
3.1.3 Complete Bloch Equations
A more technical description of the comagnetometer follows from a complete set of
Bloch equations. The evolution of the K electron spin polarization P e and the 3He
nuclear spin polarization P n follow
∂ ~P e
∂t=
γeQ(Pe)
(~Ba + λMn ~P n + ~L+ ~βe
)× ~P e − ~Ω× ~P e
+(Rp(~sp − ~P e) +Rm(~sm − ~P e) +Ren
se (~P n − ~P e)−Rsd~P e)/Q(P e) (3.2)
∂ ~P n
∂t=γn
(~Ba + λM e ~P e + ~βn
)× ~P n − ~Ω× ~P n +Rne
se (~P e − ~P n)−Rnsd~P n (3.3)
where this result is explicitly written in the rotating frame (Eq. 2.30). The electron
spins precess in the applied magnetic field ~Ba, the magnetization of the nuclear spins
λMn ~P n, lightshifts ~L, and anomalous fields coupling to electron spins ~βe. The po-
larization is also influenced by a number of rates (Table 3.1). In order, we introduce
pumping from the pump beam Rp, pumping from the probe beam Rm, spin-exchange
with nuclear spins Rense , and spin-destruction of the alkali Rsd. The terms ~sp and ~sm
explicitly account for the degree of circular polarization and orientation of the pump
and probe beams with |~s| = 1 for circularly polarized light. The slowing down factor
Q(P e) affects all the rates for the electron spins. Similarly, the nuclear spins precess
in the applied magnetic field ~Ba, the magnetization of the electron spins λM e ~P e,
and the anomalous field coupling to nuclear spins ~βn. The relaxation of the nuclear
polarization is simpler and influenced by the spin-exchange rate between nuclear and
electron spins Rnese , and the nuclear spin-destruction rate Rn
sd.
50
~P e/n Alkali/noble polarization Rp Pump beam pumping rate~L Lightshift Rm Probe beam pumping rateλM e/n Alkali/noble magnetization Ren
se Spin-exchange rate (alkali-noble)~Ω Rotation rate Rne
se Spin-exchange rate (noble-alkali)~βe/n Alkali/noble anomalous field Re
sd Spin-destruction rate (alkali)~sp Pump photon spin polarization Rn
sd Spin-destruction rate (noble)~sm Probe photon spin polarization Rtot Total alkali relaxation rateQ(P e) Slowing down factor Rn
tot Total noble relaxation rate
Table 3.1: Notation summary in the complete Bloch equations.
The terms in Eqs. 3.2 and 3.3 lead to a lot of algebra, so the appearance is
simplified with a few tasteful definitions1. Cross products are grouped and rates are
collected into a more compact form to produce
∂ ~P e
∂t=
γeQ(Pe)
(~Ba + λMn ~P n + ~L+ ~βe −
~ΩQ(P e)
γe
)× ~P e
+(Rp~sp +Rm~sm +Ren
se~P n −Rtot
~P e)/Q(P e) (3.4)
∂ ~P n
∂t=γn
(~Ba + λM e ~P e + ~βn −
~Ω
γn
)× ~P n +Rne
se~P e −Rn
tot~P n (3.5)
introducing the definitions Rtot ≡ Rp + Rm + Rense + Ree
sd and Rntot ≡ Rne
se + Rnsd.
This format highlights the similarity between the complete Bloch equations and the
simplified model presented in Section 2.5.2 that motivated the comagnetometer and
justifies the approximations made in Ref. [70]. It also highlights the similar footing
for each of the anomalous fields, proposed and practical. For reference, a summary
of the notation in the Bloch equations is available in Table 3.1.
As written, ~Ba refers to the applied magnetic fields. The behavior of interest
occurs at the compensation point (Eq. 2.67) and it is simplest to consider fields
1The work in Refs. [43, 74] choose to group and rescale these equations further. This leads to amathematically more elegant solution, but is more difficult to interpret in terms of the physicallyrelevant parameters. This author chooses to carry around the extra factors for ease of interpretationwithin the context of doing experiments related to physical parameters in the lab.
51
relative to this point. The definition,
Baz = Ba
c +Bz = −Bn −Be +Bz (3.6)
along with Bax = Bx and Ba
y = By achieves this goal. From now on, we refer to all
magnetic fields with respect to the compensation point.
3.2 Steady State Response
Searches for anomalous spin dependent forces rely on a correlated measurement with
the comagnetometer signal and the source of the anomalous field. The leading or-
der comagnetometer response is the difference in anomalous field coupling between
electron and nuclear spins. It is expected that the anomalous fields of interest are
small, so drifts from higher order corrections to the comagnetometer response become
systematic effects.
3.2.1 Leading Response
As written, this author believes the full Bloch equations to be impossible to solve
analytically. An approximate solution can be recovered by assuming that the equilib-
rium polarization for ~P e and ~P n is oriented in the z-direction. In this way, equilibrium
polarization values for P ez and P n
z can be determined
P ez =
RenseP
nz +Rm~sm · z +Rp~sp · z
Rense +Rm +Rp +Rsd
' Rp~sp · zRtot
(3.7)
P nz =
RneseP
ez
Rnese +Rn
tot
(3.8)
similar to the results obtained for spin-exchange optical pumping (Section 2.2). One
can treat the equilibrium polarizations as constants and consider small deviations in
the transverse direction as in the linearized Bloch equations (Section 2.5.2). Here, we
52
forgo the complex notation. The resulting coupled equations are linear and can be
solved exactly.
The resulting expression for P ex is similar to Eq. 2.48, but with many more terms.
With the compensation point definition enforced (Eq. 3.6), the leading terms are
P ex =
P ez γeRtot
(βey − βny + Ly +
Ωy
γn− ΩyQ
γe
)(3.9)
with magnetic field, pumping, and spin-exchange effects forming higher order terms.
These effects are difficult to observe directly from the analytic expression. The
overall expression can be separated into a numerator and denominator with terms
expanded and ranked by size2. A strategy of asserting numerical values for each term
and sorting terms in descending order has been developed in Ref. [28]. Typical values
for all of the terms in the Bloch equations are given in Table 3.2. As in previous work,
we consider the comagnetometer under normal operating conditions. The resulting
expressions are much simpler if Ly is ignored as well as βe/nx and β
e/nz ; however, these
dependencies are revisited in Section 3.2.9.
3.2.2 Numerator
From now on, we will adopt the convention of referring to P ex as the signal S. This
will be especially useful considering the dependance of S(X) on variable X. The large
number of terms in the equilibrium solution at the compensation point are broken
into a numerator N and denominator D. For simplicity, we consider
S(X) =P ez γeRtot
(N(X)
D
)(3.10)
2This involves a tour de force of algebraic trickery to coax a program such as Mathematica tocollect terms in a meaningful way.
53
Term Typical Value Term Typical Value
γn 2π×3.243 kHz/gauss Rp 166 1/sγe 2π×2.8 MHz/gauss Rm 100 1/sQ(P e) 5.2 Ren
se 36 1/sna 7× 1013/cm3 Rne
se 4× 10−6 1/snb 2.5× 1020/cm3 Rtot 353 1/sλM e −28 µgauss Rn
tot 2× 10−4 1/sλMn −130 mgauss P e
z 50%~sp 10−4, 10−4, 1 P n
z 2%
~sm 10−4, 0, 0 | ~Ω⊕| 7× 10−5 rad/s~B 10−6, 10−6, 10−6 gauss ~βe 0, 10−8, 0 gauss~L 10−7, 0, 10−7 gauss ~βn 0, 10−8, 0 gauss
Table 3.2: Estimated values for the comagnetometer under normal operating conditions.The cell is heated to 180 and filled with 9.4 amagat of 3He buffer gas. Values measuredexperimentally are within reasonable agreement of these estimates. We ignore Ly andpropagate βey and βny as the leading anomalous fields. The most relevant rotations are thoseof the Earth.
where we consider the coefficients of X in N(X) and the full denominator. A simpler
expression can be found by expanding the denominator for small values 1/(1 + ε) '
(1 − ε), but care must be taken here because (1 − ε) must be multiplied by the full
N(X, Y, Z,XY, Y Z...) to catch all of the relevant cross terms.
We will present most of our expressions with the leading P ez γe/Rtot since this term
serves as an overall measure of the scaling of the signal or sensitivity. We will group
terms in descending order of dimensionless corrections to the variable of interest.
A direct comparison of the results that follow and those presented in Ref. [28]
shows a distinct difference in the sign of the interactions. This comes from a difference
in the definition of sign of the gyromagnetic ratios in the Bloch equations and an
unaccounted for arbitrary sign inserted into the previous calculation. This can cause
confusion, so we have been explicit about sign conventions and definitions in this
work3.
3One will notice the use of Bc (Eq. 3.13) in this work for the term Bn. There is a difference of asign between the two definitions.
54
3.2.3 Denominator
The overall scaling of P ez γe/Rtot has already been removed from the signal, so we can
consider a dimensionless denominator
DS =(
1 +γ2e
R2tot
(Bz + Lz)2 + 2Cn
se + 2Bz
Bc
+ 2Dese + 2De
se
γ2e
R2tot
(Bz + Lz)2 +O(10−6)
)(3.11)
where we define Cese and De
se to simplify terms involving spin-exchange and magneti-
zations (Table 3.3). The terms in Eq. 3.11 are loosely in decreasing order except where
an algebraic regrouping simplifies the expression. The second term is roughly 10−2,
and we include terms up to 10−5. The leading terms in the denominator depend on Bz,
Lz, and spin-exchange expressions which will simplify combining the numerator and
denominator. The main dependencies of other terms will reside simply in the numer-
ator, and the presence of DS signifies that the denominator has not been expanded.
Products of fields with parameters Bc ' 3 mgauss and γe/Rtot ' 1/(20 µgauss)
that are much smaller than unity lead to expansions of the denominator. Expanded
expressions will not be valid if these conditions are exceeded. For the most part,
contributions from the denominator need not be expanded until terms involving Bz
or Lz are involved.
3.2.4 Anomalous Fields
The leading terms to the comagnetometer dependance involve the difference in anoma-
lous fields (Eq. 3.9). The higher order corrections to this response are
S(βey) =P ez γe
RtotDS
βey
(1 + 2
Bz
Bc
+ Cnse(1 + Fr) +De
se − 2Ωz
γnBc
+O(10−7)),
S(βny ) = − P ez γe
RtotDS
βny
(1 + Cn
se(1 + Fr −Bz
Bc
) +Bz
Bc
+Dese +O(10−7)
)(3.12)
55
Name Abbreviation Definition Size
Electron spin-exchange correction Cese
Pnz R
ese
P ezRtot
10−2
Nuclear spin-exchange correction Cnse
γeP ezR
nse
γnPnz Rtot
10−3
Electron spin-exchange and magnetization Dese
MeRese
MnRtot10−5
Magnetization ratio FrMeP e
z
MnP ez
10−2
Table 3.3: Dimensionless terms found in the steady state expansion containing spin-exchange and magnetization expressions. These coefficients are evaluated using the es-timates in Table 3.2.
where we introduce Fr (Table 3.3) to represent the magnetization ratios and
Bc = −Bn ' −Bn −Be (3.13)
to represent the compensation of magnetic fields by the 3He magnetization. This
choice of definition will be useful from an experimental point of view because the
applied Bc field points in the same direction as Bz. There are subtle differences
between the expressions for S(βey) and S(βny ), most notably the factor of 2 on the
Bz/Bc term and the appearance of Ωz in S(βey).
Expanding the denominator in Eq. 3.11, these expressions become
S(βey) =P ez γeRtot
βey
(1− γ2
e
R2tot
(Bz + Lz)2 − Cn
se −Dese +O(10−5)
),
S(βny ) = −Pez γeRtot
βny
(1− γ2
e
R2tot
(Bz + Lz)2 − Cn
se −Bz
Bc
−Dese +O(10−5)
)(3.14)
where we keep terms up 10−4. The leading correction is 6× 10−3 and depends on Bz
and Lz. The dependance of both expressions is similar except for the cancelation of
Bz/Bc in S(βey). If the contribution of Bz and Lz can be reduced, the response to
56
anomalous fields is simply
S(βey, βny ) =
P ez γeRtot
(βey − βny )(
1− Cnse −De
se +O(10−6)). (3.15)
3.2.5 Pumping
Before launching into several higher order terms, it is important to identify another
first order contribution. Pumping from either the pump or probe beams along the
x-direction produces
S((~sp + ~sm) · x) =Rp
Rtot
~sp · x+Rm
Rtot
~sm · x (3.16)
as an experimental imperfection. This contribution can be re-expressed in terms of
the orthogonality of the pump and probe. For an alignment angle π/2 +α, the signal
is
S(α) = αRp
Rtot
(3.17)
and generates a fixed offset in the response. Any mechanical drift through time of
the pump and probe alignment introduces a time dependant response to the signal.
Since the linearly polarized probe beam is nominally oriented along the x-direction,
we can consider its contribution in terms of the remnant circular component,
S(sm) = smRm
Rtot
. (3.18)
Once again, a contribution from this term contributes an offset and any change
through time leads directly to drift. The probe beam is nominally linearly polar-
ized, so |~sm| 1. We have implemented a scheme to reduce the remaining circular
component of the probe beam (Section 4.2.7) and to stabilize the relative orientation
of the pump and probe (Section 4.2.9).
57
3.2.6 Magnetic Fields
In a magnetometer, the leading sensitivity is to the By field. In the comagnetometer,
the leading By dependance is suppressed by a product with Bz,
S(By) =P ez γe
RtotDS
By
(Bz
Bc
− Ωz
γnBc
− Bz + LzBc
Cese +
B2z
B2c
− LzBc
Cnse +O(10−6)
). (3.19)
The leading corrections come from fields along the z-direction including lightshifts
and rotations. Each of these is suppressed by Bc (Eq. 3.13). Larger 3He polarizations
reduce the influence of these terms on the signal.
The dependance on Bz contains a dependance on By as expected from Eq. 3.19,
S(Bz) =P ez γe
RtotDS
Bz
( γeRtot
Lx(1+Fr)+By + βeyBc
−Rp~sp · yP ezRtot
+γeΩx
Rtotγn+βey − βnyBc
+O(10−6)).
(3.20)
The leading term in this case comes from Lx. Also, there is an Ωx/γn contribution
that is also modified by γe/Rtot. It is interesting to also observe the appearance
of anomalous fields βey and βny . A measurement of these effects is most direct from
Eq. 3.9. In the case that By is small, these effects are already suppressed. The appear-
ance of a pump beam projection along the y-direction also leads to a Bz dependance.
These will be particularly important since the Bz field is continuously changing due
to drifts in the polarization of 3He.
Because of the dependance of Bz in the denominator, expansion of Eq. 3.20 be-
comes
S(Bz) =P ez γeRtot
Bz
( γeRtot
Lx +By + βnyBc
− Rp~sp · yP ezRtot
+γeΩx
Rtotγn+O(10−5)
)(3.21)
where the−2Bz/Bc term in the denominator in combination with the leading (βey−βny )
response cancels much of the anomalous field response to leave a lonely βny term. This
is in agreement with Eq. 3.14. It is worth pointing out that the same cancelation
58
provides a −Ωy/(γnBc) term buried in the higher order part of the expression. This
is negligible unless large rotations about the z-axis are introduced.
Both the By and Bz fields are suppressed by the product with another small field;
however, Bx introduces a first order dependance to the signal,
S(Bx) =P ez γe
RtotDS
Bx
(γeBz(Bz + Lz)
RtotBc
+Rtot
γeBc
(Cese + Cn
se) +O(10−6)). (3.22)
where we consider the terms in the same way as S(By) and S(Bz). This can be
rearranged into the following more illuminating format
S(Bx) =P ezBx
BcDS
(γ2eBz(Bz + Lz)
R2tot
+ Cese + Cn
se
). (3.23)
The symmetry of the magnetic field cancelation is broken by spin-exchange terms
on the order of 10−2. The leading field dependance is the product of three magnetic
fields BxB2z with no binary contributions. It is also interesting to note that there is
no L2z present as there was in the denominator (Eq. 3.11).
It is useful to consider the B2z dependance since this can lead to an evaluation of
the contribution to Bx. As noted earlier, it is not simple to directly access this term,
since the denominator also carries a B2z dependance. This evaluation determines
S(B2z ) =
P ez γ
2e
R2totDS
B2z
((Bx + Lx)(1 + Fr) + LxBc
+ByRtot
B2cγe− 2Rp~sp · yBcγeP e
z
− Ωz
γnBc
+O(10−8))
(3.24)
which contains a Bx term as expected from Eq. 3.23. It is also influenced by Lx at
10−4. The effects of pumping along the y-direction and By are suppressed at 10−6.
These terms are superseded by those from the expansion of the denominator. We can
include the leading affects of the numerator by considering the product with the first
59
order terms in Eq. 3.9 to provide
S(B2z ) =
P ez γeRtot
[ γeRtot
B2z
(Bx
Bc
− γeΩy
Rtotγn+O(10−5)
)](3.25)
which picks up an important ΩyB2z term at 10−4. Also, the Lx dependance is removed
during expansion from the product of the −2Bz/Bc term in the denominator with
the leading Lx dependance in the numerator (Eq. 3.34).
Similarly, the product ByBz is worth considering as this will later provide a means
of calibrating the magnetometer (Section 4.3.1). Before expanding the denominator,
S(ByBz) =P ez γe
RtotBcDS
ByBz
[1−Ce
se−2Ωz
γnBc
− LzBc
Cese+
( Ωz
γnBc
+ΩzQ
γeBc
)Cese+O(10−11]
)(3.26)
where two Cnse terms have already canceled and the sum of effects from rotations and
from electron and nuclear spins appears. The contribution from the denominator
must be evaluated carefully by also considering the numerator with By (Eq. 3.19)
S(ByBz) =P ez γe
RtotBc
ByBz
(1− Ce
se − 2Cnse + 2
Bz + LzBc
Cese +
2Bz + LzBc
Cnse +O(10−5)
)(3.27)
which simplifies to
S(ByBz) =P ez γe
RtotBc
ByBz
(1− Ce
se − 2Cnse +O(10−5)
)' P e
z γeRtotBc
(3.28)
if the Bz and Lz fields are reduced. As defined, Bc follows the same sign convention
as Bz, so the slope reverses if the comagnetometer axes are reversed. The Ωz term
remains as a 10−5 correction.
60
3.2.7 Rotations
Sensitivity to rotations enters at the first order4 since rotations behave as anomalous
fields. The dominant sensitivity is along the y-direction,
S(Ωy) =P ez γe
RtotDS
Ωy
γn
(1− γnQ
γe+ Cn
se +Bz
Bc
+Dese +O(10−6)
). (3.29)
The terms in the numerator are similar to those in the denominator such that
S(Ωy) =P ez γeRtot
Ωy
γn
(1− γnQ
γe− γ2
e
R2tot
(Bz + Lz)2 − Cn
se −Bz
Bc
−Dese +O(10−6)
)(3.30)
with the denominator fully expanded. Rotations are dominated by the Ωy/γn term
with a 0.5% adjustment for the electron spin contribution. As with the anomalous
fields in Section 3.2.4,
S(Ωy) =P ez γeRtot
Ωy
γn
(1− γnQ
γe− Cn
se −Dese +O(10−6)
)(3.31)
for sufficiently small Bz and Lz.
Rotations about the x- and z-axes are further suppressed, but still provide some
rotational sensitivity,
S(Ωx) =P ez γe
RtotDS
Ωx
γn
(γe(Bz + Lz)
Rtot
+γeBz
Rtot
Fr −γn(Bz + Lz)
Rtot/Q+O(10−5)
)(3.32)
S(Ωz) =P ez γe
RtotDS
Ωz
γn
(− By
Bc
− γeRtot
Lx(Fr +γnQ
γe)− γeBxBz
RtotBc
+O(10−5))
(3.33)
where sensitivity to Ωz is suppressed by 10−4 whereas sensitivity Ωx is only suppressed
by 10−2. This is from the presence of γe/Rtot on terms in Ωx and Bc on terms in Ωz.
4This leads to the behavior of the comagnetometer as a sensitive gyroscope [41].
61
3.2.8 Lightshifts
Contributions to the signal from lightshifts from the probe and pump beams lead to
second order terms,
S(Lx) =P ez γe
RtotDS
Lx
( γeRtot
(Bz + Lz) +γeRtot
BzFr +2γeBz(Bz + Lz)
RtotBc
+O(10−6))
(3.34)
S(Lz) =P ez γe
RtotDS
Lz
( γeRtot
Lx −Rp~sp · yP ezRtot
+γeΩx
Rtotγn+γeBz(Bx + 2Lx)
RtotBc
+O(10−6))
(3.35)
where a common LxLz term is one of the leading effects. The numerator of Lx terms
has a strong Bz dependance whereas the Lz term has the product Bz(Bx + 2Lx).
Expansion of the denominator will completely cancel the last term in Eq. 3.34 and
the 2Lx term in Eq. 3.35.
3.2.9 Refinements
The preceding expressions are in good agreement with those presented in Ref. [28]
with the main differences including choice of sign conventions and some corrected
sign errors in this presentation. These parameters are useful for considering normal
comagnetometer operation. In this work, we introduce a third laser to introduce
additional lightshifts and pumping along the x- and y-direction, so it is useful to
include such terms in the analysis. We extend the Ly lightshift to a similar magnitude
as Lx and Lz (Table 3.2). The signal dependance becomes
S(Ly) =P ez γe
RtotDS
Ly
(1 + 2
Bz
Bc
+ Cnse(1 + Fr) +De
se − 2Ωz
γnBc
+O(10−7)). (3.36)
62
The denominator is unaffected by this change, so the full dependence can be simply
expressed as
S(Ly) =P ez γeRtot
Ly
(1− γ2
e
R2tot
(Bz + Lz)2 − Cn
se −Dese +O(10−5)
)(3.37)
where these expressions are precisely the same as those for S(βey) with the replacement
βey → Ly.
Furthermore, it is important to consider pumping effects from the pumping rate
Rl along the y-direction from a third beam
S(~sl · y) = −Rl~sl · yγeRtot
(Bz + Lz). (3.38)
If there is a misalignment into the x-direction, the signal is affected simply by Eq. 3.16.
Additional pumping along the z-direction provides no change in the signal other than
contributing to the value of the equilibrium polarization.
It is also reasonable to assume that βe/nx and β
e/nz also exist; the leading terms are
S(βex) =P ez γe
RtotDS
βex
( γeRtot
(Bz + Lz) +γeRtot
BzFr +2γeBz(Bz + Lz)
RtotBc
+O(10−6))(3.39)
S(βnx ) =P ez γe
RtotDS
βnx
(− γe(Bz + Lz)
Rtot
− γeBz
Rtot
Fr +O(10−5)
(3.40)
S(βez) =P ez γe
RtotDS
βez
( γeRtot
Lx −Rp~sp · yP ezRtot
+γeΩx
Rtotγn+γeBz(Bx + 2Lx)
RtotBc
+O(10−6))
(3.41)
S(βnz ) =P ez γe
RtotDS
βnz
(By
Bc
+γeLxRtot
Fr +γeBxBz
RtotBc
+O(10−5))
(3.42)
with no first order dependence. Since these fields are already well-constrained and
all the other fields are made small, we can neglect these effects. In the event that an
anomalous field is measured, these terms can play a role in confirming that it is real.
63
3.2.10 Typical Leading Terms
The expressions in the preceding sections cannot be blindly added together to form
one large dependence as this would double count a large number of terms. By carefully
selecting independent terms, a compact expression that incorporates the most relevant
terms relevant to experiments is
SL =P ez γeRtot
[βey − βny +
Ωy
γn− ΩyQ
γe+ Ly +
αRp + smRm
γeP ez
+Bz
(By + βnyBc
+ LxγeRtot
+γeΩx
Rtotγn
)+
γeRtot
(BxBz(Bz + Lz)
Bc
− LyγeRtot
B2z −
γeΩy
RtotγnB2z + LxLz
)]. (3.43)
where these have been selected to include the leading effects of each field. This ex-
pression can inform almost all of the measurements in Chapter 4. This is a simple
enough expression to remember the pattern after some experience with comagnetome-
ter operation. Only the Ωz term has been neglected, but this sensitivity has been
suppressed by 10−5 and will be difficult to measure unless a large rotation about the
z-axis is performed.
3.3 Experimental Implementation
The measurements in this thesis bridge two generations of comagnetometer develop-
ment. We will refer to the first generation as CPT-I and the second as CPT-II. The
comagnetometer suppresses magnetic field changes at frequencies below the com-
pensation frequency of several Hz. Therefore, long term drifts from 1/f noise can
dominate the signal in the frequency range of interest. Much of the development of
a robust apparatus addresses the issue of long term drift.
64
Polarizer
Photodiode
λ/4CellOven
Probe LaserField Coils
µ-metal Shields
Polarizer
Pump Laser
Figure 3.3: General comagnetometer apparatus including the cell, oven, shields, coil form,and optics. The comagnetometer is sensitive to anomalous fields perpendicular to the pumpand probe beams.
3.3.1 General Features
The comagnetometer cell contains several amagats of 3He buffer gas, several Torr
of N2 for quenching, and a few mg of K metal. Corning 1720 or GE 180 glass is
chosen for its chemical resistance to the alkali and its high density to prevent the
escape of 3He by diffusion through the walls. The spherical cell has a diameter
of 2.5 cm with a 2 mm wide stem from where the cell is pulled off the vacuum
manifold during the filling procedure. The cell is mounted by the stem in an oven
and heated to achieve a K vapor density of ∼ 1013/cm3. A temperature between
160-180 is maintained using a temperature sensor and simple feedback loop on the
heating elements. Nesting the oven within several layers of magnetic shielding further
suppresses external magnetic effects. Magnetic field coils within the shields provide
control of each of the independent magnetic fields and gradients. All materials inside
the shields are non-magnetic to preserve the uniformity of the local magnetic fields
(Fig. 3.3).
Two semiconductor diode lasers provide the light for independent pump and probe
beams. A simple arrangement of spherical lenses shape the beam to illuminate the
cell. A polarizer and quarter waveplate circularly polarizes the pump beam, and a
polarizer sets the initial linear polarization of the probe beam. At an alkali density of
1013 cm3 in the a 2.5 cm diameter cell, only tens of mW of pump power is required to
65
reach an alkali polarization near 50%. A final polarizer on the probe beam converts
optical rotation to intensity which is monitored on a photodiode. Achieving long term
stability for anomalous spin coupling searches is more difficult and requires a more
sophisticated design.
Magnetic shields do not reduce the sensitivity to anomalous fields. A magnetic
field exterior to the shields causes the electrons in the shield to align with the field.
The magnetic moment of the electrons anti-aligned with the spin contributes a real
magnetic field in the opposite direction. The sum of these two contributions decreases
the net effect of the applied field inside the shields. In the case of an applied ~βe, elec-
trons in the shield again align to the anomalous field and create a real magnetic field
in response. The comagnetometer experiences contributions from both the anomalous
field and the real magnetic field. Since the comagnetometer is insensitive to magnetic
fields, the magnetic field is canceled, but the anomalous field remains to generate a
response. The case of an applied ~βn is much simpler; the electrons in the shields ig-
nore the nuclear spin coupling, so an anomalous nuclear coupling passes through the
shields unimpeded. Applying a lightshift bypasses the effects of the magnetic shields
and is therefore not a useful demonstration of this argument. Similarly, the dominant
response of the comagnetometer gyroscope comes from the nuclear spin coupling [41].
A careful measurement of the electron contribution to the gyroscope response would
experimentally confirm this argument. In any event, the measurements in this thesis
present new limits on nuclear spin couplings, so the leading gyroscope response pro-
vides sufficient experimental evidence to confirm the comagnetometer’s sensitivity to
an anomalous nuclear spin coupling.
3.3.2 Historical Development
K-3He comagnetometer development began at the University of Washington under
Michael Romalis for spin-dependant tests of Lorentz violation. Specifically, interest
66
in b⊥ in the SME prompted a search for an anomalous spin coupling where the mea-
surement timescale of interest is one sidereal day. This is a long timescale for stable
comagnetometer operation. Many of the experimental challenges include stabilizing
systematic contributions to the signal without feeding back on the signal itself. This
would remove the sensitivity to the fields of interest.
The experiment was placed on a large optical table with horizontally mounted
pump and probe beams. Five-layer µ-metal magnetic shields with a 60 cm outer
(40 cm inner) diameter provided a magnetic shielding factor of 106. One of the
earliest stability problems was identified in the relative motion of the pump and
probe beams. A special nonmagnetic stainless steel optical table with a submerged
central area allowed for rigid mounting of the optics around the shields [75]. Relative
motion was monitored by raising the pumping intensity such that the sensitivity
to misalignment increases while the sensitivity to other fields decreases. Also, slow
changes in the birefringence of the cell walls induced additional optical rotation in
the probe beam unrelated to the atoms. A means of measuring the probe beam
background by blocking the pump light was developed. With the atoms depolarized,
only the background signal contributed to the measured optical rotation of the probe
beam.
The experiment was later moved to Princeton University where a complex system
of monitors and feedbacks were incorporated to stabilize the comagnetometer signal
against systematic drifts (Fig. 3.4). These improvements occurred over years with full
details in Refs. [28, 76]. Briefly, the frequency of the 1 W pump laser is monitored
with a grating spectrometer. A variable waveplate raises and lowers the intensity for
misalignment measurements as well as stabilizes the intensity. A Fabry-Perot cavity
monitors the probe laser frequency, with a barometric pressure dependence that has
been reduced using an Invar tube. The current in a tapered amplifier is adjusted to
maintain a consistent probe intensity. A Pockel cell nulls the circular polarization
67
Oven
Probe BeamOptical Table
Optical TableLower Section
Encl
osu
res
Magnet
ic S
hie
lds
Pump Beam
Pol
ariz
er
Position Detector
DeviatorPinhole
Piezomirror
Isolator
Translating Lens
Collimating Lens
(Analyzing) Polarizer
Photodiode
Position Detector
Probe BeamDiode Laser
Fab
ry-P
erot
Far
aday
Mod
ula
tor
Field Coils
l /2
y z
x
Poc
kel
Cel
l
Pinhole
Pol
ariz
er
l /4
l /2VariableWaveplate
Polarizer
Shutter
Lenses
Final Mirror
Translation Stage
Position Detector
Spectrometer
Pump Beam Diode Laser
Cell
Hot Air Flow
Collimator
InsulationCooling Jacket
Position feedback
Figure 3.4: CPT-I comagnetometer installation showing the pump and probe beam opticallayout. Lexan boxes enclose the optics to reduce the effects of air convection. ImageCredit: [28].
of the probe beam to reduce Lx and pumping effects. Despite these feedbacks, the
external cavity lasers occasionally mode hop leading to discontinuities in the signal
which are removed in the final analysis. Greater temperature stability was achieved
by stabilizing the temperature of the laser mounts to < 10 mK.
Shifts in the position of the beams are observed from long term thermal drifts of the
optical elements and the breadboard. Four quadrant photodiode position detectors
continuously monitor beam positions and provide feedback signals to piezo controlled
mirrors. Small deviations in probe beam position are particularly symptomatic due
to optical rotation from the linear dichroism of the cell. A “sweet spot” on the cell
face can be identified near the center of the cell where the leading order dependence
68
cancels. A translating lens and deviator allow for independent control of the angle
and position of the probe beam to aid in locating the “sweet spot”.
A complex series of modulation techniques were developed to reduce the second
order effects to the signal (Eq. 3.43). Within our literature, these are referred to
as “zeroing routines” and reduce the dependance of the signal on the Bz, By, Bx,
Lx, and Lz. In addition, a 3He polarization feedback has been developed. These
have been adjusted over the years, and details on the most modern implementation
of these routines are discussed in Section 4.2. A complex LabView program was
also developed to automate data acquisition and coordinate various feedbacks and
background measurements. LabView and the associated National Instruments data
acquisition cards handle all of the data acquisition. These tools were chosen such
that during a program crash, voltages on the cards remain uninterrupted such that
transient signals do not disturb the equilibrium of the spins in the comagnetometer.
A Faraday modulator made from Tb-doped glass modulates the polarization of
the probe beam at 4.8 kHz and is measured with a Standford Research Systems
SR830 DSP lock-in amplifier. The high current required in the faraday modulator
requires that it be water cooled. This provides better long term stability than a simple
polarizer or balanced polarimeter scheme and reaches a sensitivity of 10−8 rad/√
Hz.
At this level, moving air currents contribute to noise. Density variations in moving
air modulate the index of refraction leading to optical rotation. Large Lexan boxes
around the optics reduce convective air currents. All electronics with cooling fans are
placed on an equipment rack above the optical table. Furthermore, evacuated glass
tubes along the axis of the shields reduce the optical path length through atmosphere.
A large foam box surrounds the entire optical table to reduce airflow and provide
additional thermal insulation. Still, opening the door to the room introduces a wave
of air from the hallway and a noticeable shift in the signal can be observed.
69
To reduce diurnal drifts, laboratory air vents were turned off and sealed. An air
conditioner continuously runs to maintain the room at a steady temperature and move
excess heat to an adjoining room. Acoustic foam fills the gaps between the nested
magnetic shields, and a fiberglass bundle surrounds the outermost shield. Further-
more, pedestal mounts and forks are used exclusively for greater rigidity in mounting
optics. Tightening all of the table forks has led to an observable reduction in thermal
drift. A myriad of temperature sensors around the experiment record various thermal
drifts. Systematic studies of the thermal sensitivity of individual elements have been
performed by heating different elements and searching for a correlation in the signal.
Humidity sensitivity was even discovered as a result of certain epoxies.
These features improve the long-term performance of the comagnetometer and
achieve a short term sensitivity of 5 fT/√
Hz in the range 0.1-1 Hz. Long term drift
quickly dominates the signal at lower frequencies and dramatically reduces sensitivity
on the timescale of one sidereal day.
Previous Lorentz Violation Results
It was in this configuration that preliminary results on Lorentz and CPT violation
have been collected [28, 77]. These measurements spanned 15 months and contained
many days of noise compromised data. Of these, the best 120 days were selected on
the criterion that the root-mean-square noise was less than 100 fT (Fig. 3.5). Four
systematic errors were removed: (1) any correlation between the signal and another,
separate measurement, (2) purely linear or, more rarely, second-order polynomial
drifts, (3) sudden jumps in the signal associated with mode hops of the lasers, and
(4) temporary excursions from a nominal trend.
Due to large amounts of systematic drift, data in the time domain is analyzed
using a Fourier analysis method. One considers a sidereal variation in the calibrated
70
−100
−50
0
50
100
Ay
(fT
)
0 10 20 30 40 50 60 70 80 90 100 110
Day number
Ay = (0.591 ± 0.807) fT
−100
−50
0
50
100
Ax
(fT
)
Ax = (−0.760 ± 0.740) fT
Figure 3.5: Summary of amplitudes from the best 120 days from CPT-I. While the comag-netometer can achieve noise of 1 fT/
√Hz on short timescales, noise dominates the signal
on the timescale of the sidereal day. Noise in the amplitude measurements over one day ison the order of 20-30 fT. Image Credit: [77].
data with an arbitrary phase,
Ax cos (Ω⊕t) + Ay sin (Ω⊕t) + noise. (3.44)
Multiplying by sine and cosine references and averaging the data over an integral
number of periods returns one half the amplitudes Ax and Ay. The final averaged
amplitudes from this preliminary search are
Ax =(−0.76± 0.74) fT
Ay =(+0.59± 0.81) fT.
71
An adjustment to Local Sidereal Time (LST), the proper geometrical factors, and
scaling to energy units connect this result to the SME framework.
While these results are impressive, they are only comparable to the existing limit
released by the Harvard-Smithsonian group presented a few years prior to these results
[29]. Here, co-located hyperpolarized nuclear spin ensembles of 3He and 129Xe form a
3He/129Xe noble-gas spin maser. The 129Xe spin precession frequency is locked to a
hydrogen maser and the 3He is free to precess. The ratio of spin precession frequencies
is monitored and a sidereal variation in the precession of the 3He would indicate a
possible Lorentz-violating field. Over the course of a year, 90 useable sidereal day
values provide
AX = (+1.41± 1.45) fT
AY = (−0.16± 1.39) fT
where these amplitudes correspond to LST5. While the Harvard-Smithsonian results
have a slightly larger absolute uncertainty in fT, the geometry of their measurement
is better optimized for a measurement of b⊥. Using the method in section 5.4.5, the
Harvard-Smithsonian group presents a limit of bn⊥ < 1.1 10−31 GeV whereas CPT-I
presents bn⊥ < 1.4 10−31 GeV. The comagnetometer results are consistent, but not
sensitive enough to surpass the previous measurement.
CPT-I Upgrades
Several innovations were introduced to the CPT-I installation to further suppress long
term drifts [43, 77]. The lasers were replaced with distributed feedback (DFB) diodes
(Fig. 3.6). Wavelength control was achieved through temperature stabilization of the
diode rather than a macroscopic mount. A tapered amplifier on the pump beam
5This result is reported in nHz which we have translated to our preferred units of fT for a moredirect comparison.
72
provides the high pumping rates required for the misalignment measurements. The
forced hot air heating system was replaced with a high frequency ac heating scheme.
A twisted pair of high resistance wire wrapped around the oven carries ∼ 1 A of
current at 20 kHz and dramatically reduces nonuniform heating and vibrations from
airflow variations. A photo-elastic modulator (PEM) replaces the Faraday modulator
in the optical rotation measurement. This provides better performance at frequencies
lower than 1 Hz most likely due to the improved thermal stability. The Pockel cell
for removing birefringence in the probe beam is replaced with a stress plate. A piezo
stack compresses a piece of rigidly held glass to introduce a birefringence that can be
used to cancel a small circular polarization of the probe beam.
Finally, the optical path length of the pump beam is dramatically shortened using
a single mode fiber. The flap on the table bridges the gap in the depression for
the magnetic shields (Fig. 3.7). The top of the flap is well insulated by the Lexan
enclosure while air can move freely underneath. Differential heating causes the board
to flex with changing temperature. The single mode fiber reduces the net effect of
these fluctuations but still retains some sensitivity.
The upgrades improved sensitivity in the 0.1-1 Hz range (Fig. 3.8). The peak
at 11 Hz is the resonance at the compensation frequency. Below the compensation
point, the noise drops as the comagnetometer compensates for magnetic fields until
it is limited by noise from the probe beam or drift in the comagnetometer. The peaks
at 2 and 3 Hz are from mechanical resonances of the optical table and can be filtered.
Below 0.1 Hz, 1/f noise continues to dominate. The peak at 0.05 Hz comes from
the duty cycle of the cooling system. On the timescale of a sidereal day, the noise
remains hundreds to thousands of fT.
Initial return a Lorentz violation search indicated that despite the improvements
in short-term sensitivity, systematic noise on the timescale of a sidereal day had not
been significantly improved. This is a strong indicator that a new apparatus designed
73
FI
Encl
osure
s
VT
Beam ShapingLenses
Ampl
λ/
yz
x
DFB
DFB
Collimator
Oven
Field Coils
InsulationCooling JacketCell
Stress Plate
Deviator
Collimating Lens
Translating Lens
I
Pos. Det.
λ/ Pol
Piezo-actuatedMirror
FI
Beam ShapingLenses
λ/
FI
I
Shutter
PD
λ/
TranslationStage
Mag
net
ic S
hie
lds
PM
FT
ransl
atin
g Len
s
VT
VT
PEM
Pol
PD Lock-inAmplifier
BS
BS
Optical Table
Figure 3.6: CPT-I comagnetometer upgrade. PD: Photodiode, Pol: Polarizer, FI: FaradayIsolator, PEM: Photo-elastic Modulator, BS: Beam Sampler, I: Integral Feedback Con-troller, VT: Vacuum Tube. Image Credit: [43].
Lexan Enclosure
Pump Laser
Fiber
Flexing fromdifferential heating
Magnetic Shield
Table flap
y
zx
Pneumatic Vibration Isolation
Figure 3.7: CPT-I comagnetometer upgrade showing the table flap and single mode fiber.Image Credit: [43].
74
Frequency (Hz)
1
0.1
10
100
0.01
Effec
tive
Mag
net
icFie
ld A
mplitu
de
fT √H
z
0.1 1 10 100
Figure 3.8: The optimized noise spectrum from CPT-I represents an improvement from [28]in suppressing 1/f noise at low frequencies. The compensation frequency is 11 Hz. Thedramatic drop at high frequencies is from the low pass filter on the lock-in amplifier. Thepeaks at 2 and 3 Hz are mechanical resonances of the optical table. The peak at 0.05 Hz isfrom the cooling system.
from the ground up would prove more fruitful. In the meantime, the optimized noise
spectrum (Fig. 3.8) provides excellent sensitivity for a search for an anomalous spin
dependent force if a high density spin source is placed outside the outermost shield
and reversed on a timescale of several seconds. The 0.75 fT/√
Hz sensitivity achieved
at 0.18 Hz is a useful point for such a search. The details of the spin source are
discussed in Chapter 6 and the results are presented in Ref. [42]. In the meantime,
it is sufficient to discuss the development of CPT-II, where we can take advantage
of the short-term sensitivity and long-term stability of the comagnetometer for an
improved Lorentz and CPT violation search.
3.4 CPT-II Apparatus
CPT-II introduces a more compact design to reduce optical pathlengths and the
dependence on mechanical drifts (Fig. 3.9). This results in a simplified setup with
fewer feedbacks than CPT-I. All of the optics and lasers are enclosed in a 700 L bell
jar pumped down to 2 Torr of atmosphere. Evacuation of the entire optical path
75
improves thermal stability and removes effects of air convection. The pump beam is
mounted vertically such that the sensitive y-direction is in the horizontal plane. The
entire apparatus is placed on a rotating platform for reversals of the orientation on a
timescale much faster than a sidereal day. This feature serves as a lock-in amplifier
to modulate a possible Lorenz-violating field with respect to the apparatus to take
advantage of the short term sensitivity. This changes the way in which measurements
are performed.
3.4.1 CPT-II Design
A compact design reduces the possible optical pathlength for drift and also allows
for simpler temperature control. The optical layout is only a factor of 2-3 smaller
than CPT-I but is small enough to demonstrate improvement while still making use
of conventional optical components. A brief description of the features of the major
components of the new apparatus are provided here6.
Magnetic Shields
Details on the magnetic shields used in this experiment are described in Ref. [78].
Briefly, three layers of µ-metal and an inner MnZn ferrite layer provide an overall
shielding factor of 1.2× 108 in the y-direction (Fig. 3.10). The outermost shield has
a diameter of 23 cm providing almost a factor of 3 reduction in size from CPT-I. The
10 mm thick, 104 mm inner diameter ferrite has a higher electrical resistivity than
traditional µ-metal to decrease the effects of Johnson noise from thermally conductive
6It would be unfair for this author to take credit for the work in developing this apparatus.The main contributions to the development of the comagnetometer apparatus came from ThomasKornack, Sylvia Smullin, Giorgos Vasilakis, and Michael Romalis. Countless hours were spentdeveloping and improving CPT-I by Tom and Mike. Giorgos joined the team for the CPT-I upgrades.Tom, Mike, and Sylvia pioneered the design and construction of CPT-II with many consultationsfrom Giorgos. By the time that this author entered the project, the framework for an apparatuswas already assembled in the lab. The first improvements in low frequency noise from the bell jarand magnetic shielding from the ferrite shield had already been demonstrated. Modifications to theapparatus continued, but the initial development was already in place.
76
rezir
aloPME
P
edoi
doto
hPni-
kcoL
sisyl
anA
yx
z
CellOven
Probe Laser
Field CoilsFerrite
Cooling Jacket µ-metal Shields
Pre
ziral
o
Pump Laser
Vacuum Wall
OpticalIsolator
OpticalBreadboards
InductivePosition Sensors
EvacuatedBell Jar
Vibration-Isolation Stage
Electronics andEquipment Rack
Vacuum Line
Bearing
RotatingVacuum and ElectricalFeedthroughs
Encoder
4 kW ServomotorVacuumPump
0.75m
Figure 3.9: Schematic of the CPT-II comagnetometer on a rotating platform. An equipmentrack dominates the lower half of the apparatus. All of the optics and breadboards are locatedwithin a 700 L bell jar pumped out to 2 Torr of atmosphere.
electrons in the shield. The shields have < 2 cm optical access holes along three
orthogonal directions. The µ-metal shields close with a common can and lid geometry.
The ferrite is a brittle ceramic, so an overlapping lid design is impossible to machine.
The shield is in three pieces: an annulus with two opposing lids. The mating surfaces
are polished to ∼ 1 µm to complete the magnetic circuit when closed. The ferrite
is very brittle and extra care is always used when opening and closing the shields,
particularly to avoid scratching the polishes faces.
The Curie temperature of the ferrite is near 150C. A fiberglass epoxy G10 jacket
with interwoven water cooling lines maintains the ferrite at 40C even with the bell
jar evacuated and the nearby oven hot. Top and bottom water-cooled G10 plates
hold the ferrite lids in place since the polished faces can slip. Thermally conductive
gap pads between water cooling jacket and the ferrite distribute gentle pressure across
77
Figure 3.10: Left: Comagnetometer cell. Center: Boron nitride oven with the cell suspendedby the stem. Right: Shields, cooling jacket, and coil form.
the ferrite lid from brass screws holding it in place. After degaussing the shields, a
< 2 nT remnant field typically remains.
Inner Vacuum Space/Field Coils
The space between the ferrite and the oven contains a cylindrical G10 inner vacuum
space for improved thermal insolation. The lower lid is made from a PEEK high
temperature plastic, while the upper lid is made from G10 and suspends the oven
from above. Six 5/8 inch PEEK tubes extend from the inner vacuum space through
the shields and seal with anti-reflection coated windows. All joints seal with Viton
o-rings. This space is continuously pumped and maintains a pressure of 2 mTorr.
The only significant heat loss mechanism is through radiation. The inner surfaces of
the vacuum space are coated with LO/MITTM-I low emissivity paint from Solec to
further reduce heat loss.
The inner vacuum space also serves as a structure for magnetic field coils. These
apply uniform fields in each direction as well as the five independent gradients. A
home built low noise current source with a Hg battery as an internal reference provides
up to 7 mA of current for a stable compensation field. Fine adjustments to this field
78
as well as the Bx and By coils are provided by computer controlled voltages and
50 kΩ, 20 kΩ, 20 kΩ resistors respectively. The directions of the field coils have been
measured using a fluxgate magnetometer calibrated against Earth’s magnetic field to
provide a consistent check on the response of the spins. A positive voltage applied to
the By coil produces a field in the direction of the lock-in amplifier on the equipment
rack that we use as a reference. A simple switch allows us to reverse the direction
of the Bz coil. We refer to the two positions as “Forward” and “Reverse”; a positive
voltage applied to the Bz coil in the “Reverse” condition results in an upward directed
field.
Oven
A boron nitride oven houses the comagnetometer cell (Fig. 3.10). This synthetic
ceramic is chosen for its high thermal conductivity and easy machinability. The six
rectangular panels are held together with PEEK screws and nuts. Two opposite faces
have a meandering path for a twisted pair of resistive nichrome wire with double
glass silicone insulation. An aluminum nitride potting compound holds the wire in
place and provides good thermal contact to the oven housing. Circular ports allow
optical access for the pump and probe beams. As is typical for this type of design,
the faces of the cell with optical access are coldest where the cell can radiate the
most. Radiation shields project outward from the oven to limit the aperture for
radiation loss. Additional heating elements wrapped around the projections attempt
to compensate for the loss. Still, K has a tendency to condense on these regions and
slowly reduce optical transparency over a few months. One would prefer to make
the stem the coldest region of the cell; however, a separately tunable stem heater
drives the K blob to the neck of the cell to preserve the spherical symmetry for the
hyperpolarized 3He often making this the warmest region (Section 4.6.1).
79
The total resistive load of the oven is 180 Ω. Several 10 kΩ platinum RTDs provide
temperature monitoring. Many of these are left as open circuits so that no dc current
flows during operation to create a magnetic field inside the shields – particularly one
correlated with the temperature of the oven. An RTD close to the cell for temperature
feedback uses an ac excitation to continuously monitor the temperature. A home-
built, compact, audio amplifier provides ∼ 1 A of current near 300 kHz. Depending
on the load to the stem heater, this frequency is decreased because the amplifier
operates near the edge of its bandwidth. All necessary electrical connections enter
this vacuum space through a multi-pin Fischer connector in an access can connected
to one of the PEEK tubes outside the shields (Fig. 3.11).
Optical System
The pump beam enters the oven vertically, and the probe beam enters horizontally
(Fig. 3.9). This places the leading sensitivity to anomalous fields in the horizontal
plane such that a reversal of the rotating platform will reorient this axis. Two custom
304 stainless steel breadboards form the upper and lower deck of the optical layout
(Fig. 3.11). A large 12 inch circular hole extends through the lower breadboard
for placement of the shields. A 2 inch hole in the upper breadboard allows pump
light to enter the shields after reflecting off a mirror mounted at 45. The lower
breadboard is 4 inches thick for greater rigidity in optical rotation measurements
while the upper board is only 2 inches thick. These breadboards are mounted on a
mechanical vibration isolation platform with a resonance at 1 Hz since a more typical
compressed air damping system is difficult to implement within a vacuum.
Two DFB lasers from Eagleyard Photonics provide the tens of mW of light near
770 nm for the pump and probe beams. Several diodes gave their lives in the course
of the experiments here. A Zener diode and germanium diode network along with
a simple low pass filter in parallel with the diode have reduced electronic transients
80
Access Can
PEM
Photodiode
Pillars Position Detector
PositionDetector
Lens
Stress Plate
Laser
λ/2
λ/2
λ/4
λ/4
Wedge
Shields
Lower Deck: Probe Optics Upper Deck: Pump Optics
Isolator Laser
AnamorphicPrism
45o Mirror
Breakout Boxes
ShutterAperture
Wedge
Fiber
CalcitePolarizer
Figure 3.11: Optical layout for the pump and probe beams on the upper and lower decksof Fig. 3.9.
that would result in instant laser death. These lasers typically run near 773 nm at
room temperature and must be cooled to ∼ −10 to approach the D1 transition in K.
The internal thermal electric cooler (TEC) by itself does not provide enough cooling
power, so additional external TECs cool a copper plate7 mounted to the back of the
diode to ∼ 13C. Since these lasers operate in vacuum, the aluminum laser mount
carries chilled water from the same loop cooling the ferrite shield. The mounts have
been mechanically reinforced for greater rigidity under rotation of the platform. We
typically observe drifts of 10 pm/day for diodes near the end of their lifetime and
< 0.2 pm/day for well-behaved diodes.
The aspect ratio of the beams is reshaped by an anamorphic prism pair. Spherical
lenses shape the pump and probe beams to illuminate the cell. The pump beam is
expanded to beyond the size of the cell, and a rigidly mounted 15.8 mm diameter
aperture selects the spatially flat center of the beam to provide 8.3 mW of spatially
uniform pumping light to the cell. Probe light at 770.759 nm is expanded to a
6 mm beam waist to illuminate a large number of atoms, but still avoid distortion
7During the summer months in Princeton, even the external TECs operate below the dewpointcausing condensation to appear on the diode faces. Once the bell jar has been pumped out, this isno longer an issue.
81
at the edges of the spherical cell. The combination of absorption by the vapor and
condensation of metallic K on the windows decreases the 13.2 mW of light incident
on the cell to 2.4 mW.
The intensity of both beams can be controlled either by adjusting the current to
the diode or rotating a half waveplate prior to a calcite polarizer. A stepper motor
on the waveplate mount for the pump beam allows adjustments when the bell jar is
evacuated. A shutter controlled by a stepper motor can quickly block and unblock the
pump beam for measurements of pumping by the probe beam. Immediately before
entering the inner vacuum space, the pump beam passes through a quarter waveplate
to circularly polarize the beam.
The probe beam layout includes a optical polarimeter. Two crossed calcite polar-
izers on either side of the cell, a quarter waveplate, a photo-elastic modulator (PEM)
and photodiode make all of the optical rotation measurements. Measurements of
optical rotation occur between the first polarizer and the quarter waveplate (Sec-
tion 4.1.3). A stepper motor on the quarter waveplate provides fine adjustments even
when the bell jar is closed. As with CPT-I, the PEM demonstrates lower noise than
a Faraday modulator on long timescales; it is also more amenable to operating in
vacuum. Anti-reflection coated wedges sample a small fraction of each beam and
monitor the position on a four quadrant photodiode position detector.
DAQ System
A small form factor computer on the rotating platform handles all of the data acqui-
sition. Two PCI-6229 National Instruments cards read 32 analog differential input
channels and have 8 analog outputs. The computer also handles GPIB communica-
tion with the wavemeter, lock-in amplifier, and laser drivers on the platform. USB
communication with the 24-bit counter and waveplate drivers is also available. The
computer uses a wireless internet connection to communicate with the lab. Every-
82
thing on the computer can be monitored from a terminal in the lab in real time using
the built-in WindowsTMremote desktop connection.
Motor and Platform
A 4.4 kW Yaskawa ac servomotor rotates the entire 1500 lb platform. The motor is
mounted to one of the support legs 1 m from the gearbox with a ratio of 1:102. A
1 inch steel shaft transfers the rotation through a high-flex coupling. A pillow block
also supports the shaft. The original shaft was cleaved and fell to the floor while
rotating one day from a slight misalignment of the motor and repeated turning. The
coupling and pillow block have led to a reliable operation of the rotating platform.
The motor is controlled by a Yaskawa motion controller. This allows for tuning
of the PID in the servomotor and coordinated programmed moves given a velocity
profile and distance of revolution (Table 3.4). Our two common settings are “slow”
and “moderate”. A built-in filter smoothes the edges of the velocity profile and
slightly increases the time calculated to complete a rotation. Commands are sent to
the motion controller through the serial port on a computer located in the lab. A
simple LabView program translates common commands from a user interface to the
text strings understood by the motion controller. This program can repeat motions
after fixed interval delays and coordinate these moves with the computer onboard the
platform. The communication is handled through the NI-Datasocket Server over the
wireless internet connection.
In the process of finding the correct PID values for the servomotor, the platform
was occasionally set into large oscillations. Several large stop buttons around the lab
cut power to the motor for a hard emergency stop. The motor will suddenly lock and
the experiment spins 10-15 across the floor. We considered bolting the experiment
to the floor, but have chosen to let the energy dissipate through friction with the floor
83
Slow ModerateAcceleration 0.15 rad/s2 0.38 rad/s2
Deceleration 0.15 rad/s2 0.38 rad/s2
Max Speed 0.60 rad/s 1.20 rad/sReversal Time 10.4 s 5.2 s
Table 3.4: Platform reversal specifications at each of the two common speeds.
rather than in the gearbox or shaft. Service to the worm gear, shaft, or motor would
cause serious delays to the project.
The central axis of the platform carries a vacuum line as well as the 60 Hz ac
line power to the platform. A rotating vacuum seal and an electrical slip ring allow
continuous rotations of the apparatus in either direction. The earlier Huntington Labs
differentially pumped rotating vacuum seal was replaced when the o-rings failed. It
has been upgraded to a Ferrotec 1 inch hollow shaft ferrofluid rotating seal. The
ferrofluid o-rings are expected to provide a longer working lifetime than the previous
implementation. We have enjoyed two years of operation without trouble from this
device.
Around the vacuum seal, we have installed a incremental rotary encoder from
Gurley Instruments. The 1 inch hollow shaft encoder fits precisely around the vacuum
tube. A simple quadrature counter from J-Works reads the quadrature output and a
simple LabView program on the platform translates the number to a rotation angle
of the platform. The 11250 counts on the disc, four read heads, and an interpolating
function can distinguish rotations from translations to provide a precision of 0.001 in
the relative angle between the base and platform. This incremental encoder survived
most of the Lorentz violation measurements but currently loses a large number of
counts when rotating through one quadrant, probably from mechanical misalignment
of the disc and heads. The number of counts lost each revolution varies, so the encoder
quickly loses position. Reliable operation over several months has been achieved, but
future experiments should consider a more robust solution.
84
Bell Jar
The 700 L, 304 stainless steel bell jar encloses all of the optics and lasers. It has a
3/16 inch wall thickness and 34-7/8 inch inner diameter. An I-beam and hand winch
in the ceiling rated to 2000 lb can raise and lower the 300 lb bell jar. This winch
can also lift the entire experiment from the eyebolts in the 1 inch thick aluminum
baseplate. The baseplate is 36 inches in diameter and passes all of the electrical,
vacuum, cooling water, and optical fiber feedthroughs. A large diameter Viton o-ring
makes a tight seal when the bell jar rests flat against the o-ring on a bed of clean
vacuum grease.
A large number of electrical connections are supplied by 3 sets of 50-pin ribbon
cables. The ribbon cables pick up excess 60 Hz noise from the room, but simply
wrapping them in aluminum foil, insulting them with tape, and grounding one end
reduces the pickup. Twisted pair ribbon cable would prevent possible crosstalk, but
this has not been an observed issue. Several Fischer connectors and BNC connections
provide feedthroughs for more specific channels.
Once the bell jar is lowered, there is no longer optical access to the interior. It was
earlier proposed to add viewports to optically monitor the position of the vibration
isolation platform [79]. Lighting levels around the room differ and there was concern
that correlations with reversals may appear on photodiode monitors. No optical
access was added and monitoring of the vibration isolation platform with respect to
the baseplate is achieved using two non-contact inductive position sensors on either
side of the stage (Fig. 3.9).
A dry roots pump at 9 cfm evacuates the bell jar below 1 Torr within 30 min-
utes. Off-gasing from plastics and epoxies within the bell jar is expected, though this
vacuum is well-maintained over weeks. The large volume results in low sensitivity
to off-gassing with only an increase to 4 Torr observed after a few weeks. At this
point, experiments are simply paused to pump down to 1 Torr once again. When
85
0.01 0.1 1 10 1001
10
100
103
104
Frequency (Hz)
Lase
rPosi
tion
Nois
e(n
m/√
Hz)
Bell Jar Suspended
Lower Bell Jar
1 Torr Air
170 Torr He
170 Torr He Settled
Figure 3.12: Decrease in the motion of the the laser positions with removal of air fromthe bell jar. A low pass filter with 3 dB point near 10 Hz accounts for the roll-off at highfrequencies until it is limited by the bit noise from the DAQ. A repair to the water coolinglines improves the thermal stability such that the motion at 1 Torr of air improves uponthe performance of the settled He in this study.
not evacuating the bell jar, the dry pump backs a Molecular Drag Pump (MDP) that
continously pumps on the inner vacuum space necessary to maintain the 2 mTorr
inner vacuum. This area is heated to 185C and continuously offgasses. If pumping
stops, the pressure will reach hundreds of mTorr within 20 minutes failing to insulate
the oven as evidenced by rising temperatures around the coil form and ferrite.
The bell jar reduces the effects of air convection on the position of the laser
beams (Fig. 3.12). Without the bell jar, the root-mean-square motion is hundreds
of nanometers. Simply lowering the bell jar reduces these effects to < 10 nm with a
knee just below 10 Hz. Pumping out the air further improves the motion but is then
limited by thermal drift. At the time of this comparison, water cooling on the lasers
had a leak and was disabled. To solve the thermal equilibration issue, the bell jar can
be filled with 4He. He has an factor of six larger thermal conductivity and a factor of
ten smaller index of refraction with respect to vacuum than air. Allowing the He to
86
settle and equilibrate over a few minutes greatly reduces the drift. In the time since
this comparison, the cooling issues have been solved. A 1 Torr vacuum performs even
better than the He since its index of refraction is even closer to that of vacuum.
Equipment Rack
The equipment rack has a 25 inch×25 inch aluminum frame and stands 36 inches
tall. This size provides a compact design, but provides ample space for key pieces of
commercial equipment such as the lock-in amplifier, computer, and laser drivers. In
addition, a small frame 300 W chiller supplies cooling water to the laser mounts and
ferrite. Custom electronics have been have been designed in some instances for a more
compact design including analog position detector circuits, a temperature controller
feedback, and an audio amplifier. All of the equipment related to comagnetometer
operation except vacuum pumps resides on the platform to allow continuous rotations
without a large number of rotating feedthroughs.
3.4.2 CPT-II Performance
The improvements of the bell jar and compact design provide suppression of long
term drifts by almost an order of magnitude in frequency over CPT-II (Fig. 3.15).
The peak at 2.5 Hz is a mechanical vibration while the peak at 5 Hz appears suspi-
ciously like a second harmonic. Only the first peak is observed on an accelerometer.
The compensation frequency here is at 10 Hz. Below the compensation point, the
sensitivity is limited by the white electronic noise of the photodiode amplifier and
lock-in amplifier system to an effective magnetic field of 0.7 fT/√
Hz. Below 0.05 Hz,
the probe beam is limited by drift. Below 0.03 Hz, the comagnetometer is no longer
limited by the probe beam, but begins to drift as well8. The noise at 1/sd is still large,
but the comagnetometer achieves an effective sensitivity of 2 fT/√
Hz at 0.023 Hz.
8This is most likely from a drifting Bz field as the 3He polarization drifts away from the com-pensation point without feedback.
87
Figure 3.13: Image of the CPT-II apparatus on a rotating platform. The lower half isdedicated to an equipment rack. A 700 L stainless steel bell jar surrounds the optics andlasers. A view of the interior of the bell jar showing the optics and vibration isolationplatform has been superimposed on top of this image. A platform around the experimentsimplifies access to the optics when the bell jar is raised. The vacuum pumps are seen atthe bottom in the foreground. The shaft for the motor can be seen on the leftmost leg.Large Helmholtz coils surround the experiment to reduce Earth’s magnetic field. The coilsare supported by inexpensive, but stable, “mounting” cans.
88
Lock-in Side +90˚ +270˚+180˚
Figure 3.14: The four faces of CPT-II. The SRS lock-in amplifier on the lower equipmentrack provides a handy reference. Using a right handed spin-up convention, we rotate theapparatus to view all four sides.
0.1
1
10
100
Frequency (Hz)
Effec
tivve
Mag
net
icFie
ld A
mplitu
de
Comagnetometer
Probe Beam
fT √H
z(
)
0.01 0.1 1 10 100
Figure 3.15: The optimized noise spectrum from CPT-II represents a dramatic improvementin suppression of 1/f noise. The compensation frequency is 10 Hz. The peak at 2.5 Hz isa mechanical vibration. The peak at 5 Hz appears to be a higher harmonic that does notappear on an accelerometer.
89
This timescale is short enough to provide both low noise and long enough to
reverse the orientation of the comagnetometer. The motor is capable of executing
a reversal in 6 s, however it is important to let the vibration isolation platform and
the spins damp for several seconds before recording the signal. Rotations every 22 s
allow ample time for reversal, damping, and 3-7 s of signal averaging before the next
reversal.
The spectrum in Fig. 3.15 was recorded with the platform at rest over 900 s. Rota-
tions of the platform are expected to disturb the equilibrium and introduce additional
drift. This additional drift is much smaller than the drift over a sidereal day, so mod-
ulating the direction of the experiment improves the measurement methodology.
Noise Optimization
Optimizing low frequency noise takes substantial effort and patience. Often, it is
unclear where the drift originates. A correlation analysis with the signal and other
parameters can reveal the source of the drift. The comagnetometer signal has contri-
butions of drift from the probe beam and the atoms. Separation of these two signals
can help identify the source of the drift as presented in Fig. 3.15. The comagnetome-
ter can best be measured when the 3He polarization has reached a stable equilibrium
over many hours. This allows the compensation point to be maintained without an
active feedback. The probe signal can be distinguished by increasing the compensa-
tion field by several mgauss and blocking the pump beam. This depolarizes the alkali
atoms and pins the spins along the direction of the field to desensitize the response
from the atoms. Typically, the comagnetometer spectrum is taken first and followed
by the spectrum of the probe beam alone. It takes many hours to re-equilibrate the
3He polarization after depolarizing the alkali atoms by blocking the pump beam.
90
3.4.3 An Improved Lorentz Violation Search
The noise of the comagnetometer on the timescale of 100 s is on the order of the
limit set by the Harvard-Smithsonian group and CPT-I (Section 3.3.2). One would
expect that simply reversing the experiment in an arbitrary direction multiple times
could quickly average down to below the sensitivity of the previous results. A simple
“string analysis” technique (Appendix E) can determine the reversal-correlated am-
plitude response and reject any remaining longer-term drift. However, the rotating
platform introduces systematic effects that must be accounted for; any correlation
with a reversal results in a systematic effect.
The most dominant systematic effect comes from the comagnetometer’s sensitivity
to rotations as a gyroscope. In CPT-I, a response from Earth’s rotation enters as a
stable offset to the signal. Here, the response is correlated with reorientation of the
y-direction of the experiment. It achieves maximal overlap with the Earth’s rotation
axis along the N-S direction in the lab (Fig. 3.16). Based on the latitude of the lab
in Princeton, NJ, the expected amplitude of response from Eq. 3.31 is
Ω⊕γn
(1− γnQ
γe
)cos (40.345) = 271 fT (3.45)
where Ω⊕ = 2π/(86164.09 s) to account for both the rotation of the Earth and its
orbital motion about the sun with respect to a celestial inertial frame (Appendix B).
The large gyroscope response breaks the symmetry of reversals in the lab and
creates a natural basis for a Lorentz violation test. Since the Earth’s rotation rate is
stable to better than a part in 10−8, repeated measurements along either the N-S or
E-W direction should provide a stable background response. A sinusoidal variation
at a sidereal frequency in either Earth-fixed response would provide evidence for a
possible Lorentz-violating field. The large gyroscope background limits the experi-
ment to operate in an Earth-fixed frame. Measurements over the course of an entire
91
0 100 200 300 400 500 600 700 800-300
-200
-100
0
100
200
300
Heading (degrees)
Effec
tive
Magn
etic
Res
ponse
(fT
)
East - West North - South
x
x x
x
Figure 3.16: Earth’s rotation as a function of heading of the comagnetometer. Each pointrepresents 8 reversals of the apparatus. Error bars are smaller than the size of each pointand are omitted. The curve is a fit with an amplitude of 277 fT and phase of 1.1W of theexpected direction. Rotation conventions are discussed in Section 4.1.1.
day are still required to search for a sidereal variation in the reversal-correlated am-
plitude. However, the reversals of the platform allow repeated measurements with
the 2 fT/√
Hz sensitivity in the short term behavior, a significant improvement over
CPT-I. However, changes in the measured gyroscope response can limit this tech-
nique.
Reversals along the E-W axis have the greatest sensitivity to measurements of b⊥
but are most sensitive to accurate reorientation of the platform. A 0.2 variation in
the positioning of the platform leads to 1 fT noise along the E-W axis. This is a second
order effect for measurements along the N-S axis and therefore further suppressed.
Since there is a large offset in the N-S response, imbalanced changes in sensitivity or
the calibration of the comagnetometer lead to variations in the signal. A 0.3% change
in the sensitivity not captured by the calibration produces a 1 fT change in the N-S
response. Furthermore, the sensitivity of N-S measurements to b⊥ is suppressed by
sin (40.345).
92
272
274
276
278
280
282
N-S
Rev
ersa
ls (f
T)
3670 3672 3674 3676 3678 3680 3682 3684 3686 3688 3690
10
15
20
25
Sidereal Day from Jan 1, 2000
E-W
Rev
ersa
ls (f
T)
Figure 3.17: A typical long measurement run of reversals along both the N-S and E-Wdirections. The scatter is 2-3 fT which is already on the order of the previous limit [29].
In this work, we choose to make measurements along both the N-S and E-W direc-
tions throughout the course of the day. This allows for discrimination of systematic
effects along each measurement axis from changes in positioning of the platform as
well as sensitivity calibrations over the course of a day. Furthermore, a real Lorentz-
violating response would appear out-of-phase in the response of both orientations
which will aid in further distinguishing a real signal from systematic effects.
Typical long term measurements of both the N-S reversals and E-W reversals
have a scatter of 2-3 fT over the course of a sidereal day, and this performance
can be maintained over weeks at a time (Fig 3.17). It is clear that a possible sidereal
amplitude must be smaller than this scatter which is already a dramatic improvement
over the previous results (Fig. 3.5). In one day of measurements, the sidereal variation
is already on the order of the previous limit at 1 fT. However, several additional
systematic effects must be addressed before discussing the full details on the Lorentz
93
violation search. The earliest measurements of the gyroscope response resulted in
a signal facing 20 W of the local N-S axis and amplitudes 15-30% too large. This
launched a detailed investigation of the behavior of the comagnetometer gyroscope
as well as optical rotation measurements on the rotating platform. These details are
discussed in Chapter 4 as they influence the Lorentz violation results in Chapter 5.
94
Chapter 4
Rotating Comagnetometer
Measurements
The CPT-II apparatus on a rotating platform introduces a new measurement scheme
for an improved Lorentz violation search. Much of the underlying comagnetometer
operation is unchanged. Platform reversals provide a clean measurement of the ro-
tation of the Earth from the gyroscope response of the comagnetometer. The total
response of a reversal measurement can also be influenced by background effects cor-
related with other local anisotropies. This chapter describes operation of the K-3He
comagnetometer in the CPT-II apparatus and highlights new effects from platform
reversals.
4.1 Reversal Measurements
The noise spectrum in Fig. 3.15 demonstrates that measurements made at 0.023 Hz
can take advantage of the short term sensitivity of the comagnetometer. A typical
reversal measurement interval consists of continuous recording of the signal for 8-10
reversals every 22 s. Approximately every 200 s, acquisition is paused to adjust the
compensation field from drifts in the 3He polarization (Section 4.2). The first 3-7 s
95
contributes to a measurement while the apparatus is stationary. The platform is then
reversed over 6 s. Before the next 3-7 s of signal acquisition, several seconds pass for
damping of the spins and the vibration isolation stage. Every time the motor begins
to move, a trigger timestamp is saved to a file. The clocks on the two computers are
synchronized to a navy timeserver with variations less than 20 ms between them. This
allows for consistent selection of the signal with the experiment at rest prior to the
trigger. The reversal correlated amplitude is then determined using an overlapping
3-point string analysis (Section E.3).
4.1.1 Rotation Conventions
A positive voltage applied to the By coil points the same direction as the lock-in
amplifier on the platform. We use this side of the platform as a reference (Fig. 3.13).
Typically, we start measurements with reference to the local south position1 facing
the computer terminal (Fig. B.5). Rotations are most often in the “spin up” direc-
tion according to the right hand rule convention, but continuous rotations in either
direction are allowed. Each reversal rotates in the same direction to avoid the 0.2 of
backlash in the worm gear.
We define the heading of the comagnetometer as the direction in which the lock-
in amplifier points. We use the common convention of degrees measured clockwise
from True North. If the lock-in is pointing east, the heading is 90. This is opposite
from the typical direction of applied rotations (and increasing counts on the encoder)
but simplifies matters relating to the direction of the gyroscope response. For ease
of visualization, we extend headings beyond values of 0 − 360 to indicate repeated
measurements. The orientation of the comagnetometer is simply the heading modulo
360.
1See Appendix B for a determination of True North with respect to the lab.
96
The sign of the the comagnetometer response to fields can be reversed by reversing
the direction of the spins. As described in Section 3.4.1, the Bz coil points up in the
“Reverse” position. In the descriptions that follow, we will refer the reader to the
correct coil orientation and heading where relevant. A proper calibration (Section 4.3)
will clarify this ambiguity.
4.1.2 Gyroscope Measurements
Each of the points in the gyroscope map in Fig. 3.16 represents 8 measurements
from 7 reversals of the platform. A typical interval for such a measurement is in
Fig. 4.1a where the raw signal has already been multiplied by an overall calibration
value (Section 4.3). The applied rotation introduces a large response every time the
platform reverses. Structure in this response comes from motion of the vibration
isolation platform and imperfections in the platform bearings. The signal quickly
damps to the steady state value after the platform comes to rest. The overall offset
of the signal is arbitrary and comes from the relative angle of the polarization optics
in the polarimeter (Section 4.1.3).
The signal at rest is averaged to a central value (Fig. 4.1b). The uncertainty of each
point is determined from the standard deviation of the signal before averaging and the
effective noise bandwidth of the measurement (Section E.2). A 3-point overlapping
string analysis determines the reversal correlated amplitude and removes linear drift
between reversals along the entire measurement interval. The uncertainties in each
string point are much larger than the scatter of successive string points due to the
correlation between overlapping strings (Fig. 4.1c). These string points are averaged
and the standard error determined with a correction for non-independent points. The
first point is often distinct from the average of the rest of the points due to a change
in the background drift at the start of regular rotations. Prior to each interval, the
apparatus remains at rest for 30-60 s of feedback and calibration routines.
97
0 20 40 60 80 100 120 140 160 180 200
−20
−10
0
10
20
30
40
50
Time (s)
Sig
nal
Res
pon
se (
pT
)
0 50 100 150 200
6.0
6.2
6.4
6.6
6.8
7.0
Time (s)
Sig
nal
Res
pon
se (
pT
)
0 2 4 6 8 10−276.5
−276.0
−275.5
−275.0
−274.5
−274.0
−273.5
−273.0
Str
ing
Am
plitu
de
(fT
)
Index
a)
c)b)
Figure 4.1: Typical rotation interval for measurements a) Calibrated signal with largetransients from rotations. b) Data is selected while the apparatus is stationary (yellow). c)3-point overlapping string analysis of selections.
For this particular interval, a value of -276.4 fT with a statistical uncertainty of
0.4 fT is measured. The negative amplitude confirms that the measurement began
in the south position. These measurements can be repeated along different headings
to produce the points in Fig. 3.16. The statistical uncertainty in each measurement
is smaller than the size of the displayed point and is typically omitted from plots for
ease of interpretation.
4.1.3 Optical Rotation Measurements
Optical rotation measurements in CPT-II use a PEM to modulate the polarization of
the light to separate the signal from 1/f noise (Fig. 4.2). The first calcite polarizer
determines the initial linear polarization axis. Faraday rotation from the atoms ro-
tates the plane of polarization by an angle θ. Next, a quarter waveplate with one of
98
PolarizerPEM
PhotodiodeLock-in
Analysis
λ/4
Cell
Probe Laser
Polarizer
Figure 4.2: Optical polarimetry measurements using a PEM.
its axes nominally aligned with the calcite polarizer contributes a circular component
proportional to θ. A Hinds Instruments PEM with its modulation axis oriented at
45 modulates a bar of fused silica at ω = 2π× 50 kHz. The mechanical stress in the
crystal introduces a time varying birefringence with a retardation angle α = 0.08 rad.
A final calcite polarizer crossed with the original polarizer selects the time varying
intensity of the transmitted light.
The evolution of the polarization can be predicted using the Jones Matrices [80].
The PEM is modeled as a time-dependant quarter waveplate and the cell modeled as
a half waveplate. A rotation matrix determines arbitrary orientations of each optical
element. In the case that the polarizers are well aligned and the cell creates an optical
rotation angle θ, the time-varying intensity on the photodiode is given by
I(θ, t) = I0 sin(θ +
1
2α sin (ωt)
)' I0αθ sin (ωt) +
I0α2
4sin2 (ωt) (4.1)
where I0 is the initial intensity, and we have made made the approximation that
θ, α 1. In the case that θ = 0, the second harmonic response dominates the
intensity (Fig. 4.3). The first harmonic provides a signal proportional to the optical
rotation from the cell and is measured by lock-in amplifier referenced to the PEM.
Nominal alignment leads to the behavior described in Eq. 4.1, but it is interesting
to consider the deviations from this case. The individual and combined behaviors
can be easily confirmed with numerical simulations. These simulations have been
99
0 5 10 15 20 25 30 35 400
0.2
0.4
0.6
0.8
1.0
1.2
1.4
Time (µs)
Phot
odio
de
Res
pon
se (
V)
θ = 0θ
≠ 0θ
≠ 0, dc offset
Figure 4.3: Photodiode response for the polarimetry measurement for three simulated cases.
separated into interesting cases to produce simple analytic expressions. If the quarter
waveplate is rotated by an angle χ
I(θ, χ, t) ' I0(2χ2−2χθ+θ2)+I0α(θ−χ) sin(ωt)+I0α
2
4(1−2θ2−4χ2−4θχ) sin2(ωt)
(4.2)
and the first harmonic responds in the same way as optical rotation from the atoms.
For a small adjustment δ to the final polarizer,
I(θ, δ, t) ' I0(δ2 + θ2) + I0α(θ − 2δ2) sin(ωt) +I0α
2
4(1− 2δ2 − 2θ2) sin2(ωt) (4.3)
leads only to a second order correction in δ to the first harmonic response. Similarly,
a rotation of the PEM by an angle ζ produces only a second order correction to the
optical rotation signal
I(θ, ζ, t) ' I0θ2 + I0αθ(1− ζ2) sin(ωt) +
I0α2
4(1− 2θ2 − 2ζ2) sin2(ωt). (4.4)
We can model the addition of circular birefringence to the cell or intervening optics by
considering a quarter waveplate with its fast axis nominally aligned with the incident
100
polarization. For a small rotation φ of this waveplate,
I(θ, φ, t) ' I0(θ2 +φ2
4) + I0αθ sin(ωt) +
I0α2
4(1− 2θ2 − φ2
2) sin2(ωt) (4.5)
and there are no corrections to the first harmonic. Of all these possible misalign-
ments, the leading change to the first harmonic response comes from a rotation of the
first polarizer and the quarter waveplate as these define the axis of optical rotation.
Therefore, these optics must be rigidly mounted to avoid undesired changes in the
optical rotation signal.
Each of these misalignments to the nominal case leads to a dc offset in the pho-
todiode response (Fig. 4.3). In principle, this effect can be removed by adjusting
optical elements until the optimal configuration can be found (Eq. 4.1). For example,
a remnant birefringence can be removed by adjusting the final polarizer
I(θ, φ, δ, t) ' I0(δ− φ
2)2 + I0αθ(1− 2δ2) sin(ωt) +
I0α2
4(1− 2δ2− φ2
2) sin2(ωt), (4.6)
however, experimentally this is not the case. A minimum for the offset can be found
that is difficult to reduce below 5% of the total response. Simulations indicate that a
number of configurations should lead to a null offset though they neglect any spatial
non-uniformity in optical rotation or birefringence. Simply, if there is a different
response on two halves of the beam profile, an adjustment of the final polarizer can
only compensate for one half at a time. The sum of the two signals leads directly to
an offset in the presence of a non-uniformity. This offset is mostly an experimental
curiosity and does not contribute significantly to the measurement itself.
In the optical rotation measurements, care is taken to remove extra elements from
the optical rotation axis such as dielectric mirrors. Mirrors would yield greater flexi-
bility in beam alignment, but the difference in reflectivity of different polarizations can
influence measurements. Measurements reach a noise noise floor of 4×10−8 rad/√
Hz.
101
0 20 40 60 80 100 120−300
−200
−100
0
100
200
300
400
500
Time (s)
Pro
be
Pos
itio
n (
nm
)
FloorLevel
Figure 4.4: Deflection of the probe beam horizontal position measured on a four quadrantphotodiode during a typical reversal interval. The 200 nm spikes occur while the experimentis accelerating. The position settles to a fixed value at rest. By raising the legs of theexperiment off the floor, the breadboards can be leveled to reduce the shifts in positionobserved at rest.
Frequencies above 0.1 Hz reach the floor whereas longer term signals are limited by
drifts and 1/f noise.
4.1.4 Tilt
Prior to this work, it was reported that the lasers do not change position under
rotation of the platform [79]. This claim came from position sensors whose effective
sensitivity had been reduced by placing the detector at the focal point of a lens. It
is common to sacrifice sensitivity by reducing the beam size to fit within the 1 cm
diameter detector face using a lens. At the focal point of the lens, the decrease in
sensitivity to translations of the beam is almost infinite.
As initially constructed, the rotation axis leans 0.5 from vertical. Given this
anisotropy, the apparatus experiences changes in gravitational field in different posi-
tions. The probe beam breadboard is designed to be monolithic, but shifts in beam
102
−0.4 −0.3 −0.2 −0.1 0 0.1 0.2 0.3 0.4 0.50.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
Tilt Magnitude (degrees)
Optica
l R
otat
ion (μr
ad)
Figure 4.5: Typical optical rotation signal as a function of the magnitude of the tilt. Nega-tive tilt values indicate that the apparatus has been tilted beyond vertical. Optical rotationis not completely removed when leveled because other effects persist.
position on the order of 50 nm are observed (Fig. 4.4). Larger deflections occur under
acceleration of the platform but return to a consistent value at rest.
The tilt is monitored using two orthogonally placed C-29 tilt sensors from Ad-
vanced Orientation Systems Inc. (AOSI) mounted near the center of the upper bread-
board. These provide 1 µrad sensitivity over a range of 3 degrees with a bandwidth
of 2 Hz. Simply elevating two legs of the apparatus ∼ 1 cm orients the rotation axis
within 0.005 of vertical and suppresses the correlations at rest. For a typical po-
larimeter arrangement, the optical rotation as a function of tilt angle is 1.4 µrad/deg
(Fig. 4.5). It is interesting that the optical rotation magnitude is not symmetric about
the vertical. This is likely due to a redistribution of weight on the vibration isolation
platform.
A series of measurements to determine the source of the beam walk have been
undertaken. The response persists for even a minimal polarimeter on the edge of the
breadboard with only the necessary optical elements. Several polarizers, mounts, and
waveplates have been tightened, loosened, and swaped with no qualitative difference
in response. Gentle pressure on laser mounts leads to an optical rotation response.
103
These mounts have been mechanically reinforced and the effect reduced. Stressing the
breadboard near polarization optics by hand can also induce a small optical rotation
response. No particularly offensive optic has yet been identified.
A 50 nm shift in position of a 5 cm tall waveplate mount with respect to the base
could appear as a 1 µrad rotation of the waveplate (Eq. 4.2). This is smaller than
the observed effects before leveling the apparatus. Imperfections in the birefringence
of the optics along different beampaths likely leads to an enhancement of this effect.
Capacitive coupling of the PEM modulation frequency onto the photodiode amplifier
has been observed as an additional tilt effect. This and other effects have been
removed by shielding and strain relieving all cables.
4.1.5 Faraday Effect
The Faraday effect describes the rotation angle θ in the plane of polarization of linearly
polarized light as it passes through a medium with a magnetic field B parallel to the
direction of propagation. The rotation is given by
θ = V Bl (4.7)
where V is the Verdet constant of the material and l is the propagation length through
the medium. The Verdet constant can vary with temperature and wavelength.
Earth’s magnetic field has a preferred direction in the lab. The magnetic shields
and magnetic field compensation of the comagnetometer are enough to suppress
Earth-field size effects on the atoms to below 0.1 fT. A reversal of the apparatus
reverses the external magnetic field and the contribution of the Faraday rotation to
the optical rotation signal. The leading optical contributors to the Faraday effect are
the polarizer, stress plate, vacuum windows, lens, and quarter waveplate (Fig 3.11).
The Verdet constants and thickness of each element is listed in Table 4.1.
104
Optic Material Verdet Constant Thickness Reference(µrad/gauss·cm) (cm)
Polarizer Calcite 5 1 [81]Window, Lens BK7 glass 4 1, 0.3 [82]Waveplate Quartz 5 <0.01 [80]Stress Plate Soda Lime Glass 3 0.1 [83]
Table 4.1: Verdet constants of optics in the polarimeter.
Large 2.4 m square Helmholtz coils surround the experiment to reduce the changes
in magnetic field under reversals to below 5 mgauss. A three-axis fluxgate magne-
tometer on the lower breadboard monitors all three components of the magnetic field
during measurements. Based on the Verdet constants in Table 4.1, an optical rotation
amplitude of 4 µrad is predicted from the measured 0.44 gauss magnetic field. The
observed effect is 11.2 µrad and is well-aligned with the field direction (Fig. 4.6). The
high permeability shields concentrate the field in the region of the optics and the
fluxgate is located 15 cm from the magnetic shields. This factor of 2-3 enhancement
is enough to account for the discrepancy.
4.1.6 Optical Interference
A scan of the probe beam wavelength produces oscillations in the optical rotation
signal characteristic of an interference pattern. Sensitivity to motion of the optical
components has already been identified on the scale of 50 nm (Section 4.1.4) and
these shifts in pathlength can likewise influence the phase of the interference pattern
under reversals of the platform (Fig. 4.7).
Observed interference fringes vary the optical rotation response from 5-200 µrad.
The source of the pattern is expected to depend on the length of the cavity causing
the interference. In Fig. 4.7, the high frequency fringes indicate a cavity on the scale
of 0.5 m whereas the lower frequency gives a cavity length of 5 cm. These lengths
do not correspond to any particular optic or spacing between optics. Modifications
105
−15
−10
−5
0
5
10
15O
ptica
l R
otat
ion (μr
ad)
0 100 200 300 400 500 600 700−0.50
−0.25
0
0.25
0.50
Flu
xga
te R
espon
se (
gauss
)
Relative Heading from Applied Field (degrees)
a)
b)
Figure 4.6: a) Optical rotation from reversals in an applied magnetic field. b) Fluxgateresponse measured along the probe beam propagation direction. Both responses have thesame phase.
−5 0 5 1075
80
85
90
95
100
105
110
Optica
l R
otat
ion (μr
ad)
Probe Wavelength Scan (pm)
westeast
Figure 4.7: A slow scan of the probe laser wavelength produces an interference pattern inthe optical rotation signal of the probe beam. The phase of the pattern can change withreversal of the apparatus.
106
−400
−300
−200
−100
0
100
200
300
400
Time (s)
Sig
nal
(fT
)
20 13411596775839 191172153
Signal
BackgroundSelection
Figure 4.8: Reversal measurements are modified to include a probe background measure-ment in each position of the platform. The difference between the signal selections (red)and the background (green) is processed in the string analysis.
to the optical arrangement can dramatically change the frequency of the observed
pattern, but only in a seemingly unpredictable way. The fringes persist even when
the experiment is well-leveled and have a relatively stable phase over short timescales.
4.1.7 Optical Rotation Background Measurements
Reversal correlated changes in the probe beam from the beam walk and Faraday ef-
fect can be reduced to 100 nrad by controlling the tilt and the local magnetic field.
To remove residual probe background contributions from our signal, a background
monitoring scheme has been developed. A relatively large Bz field of 0.65 mgauss de-
sensitizes the atoms and provides a real time measurement of the probe background.
In each position, the Bz field can be smoothly raised for a correlated measurement of
the background for 2 s before returning it to the compensation point for a measure-
ment of the comagnetometer response (Fig 4.8).
This scheme was implemented for the second half of the Lorentz violation mea-
surements in Chapter 5. The long term behavior of the the probe beam background
107
270
275
280
285
N-S
Rev
ersa
ls (
fT)
3669 3670 3671 3672 3673 3674 3675 3676 3677 3678270
275
280
285
Sidereal Day
N-S
Rev
ersa
ls (
fT)
a)
b)
Figure 4.9: a) N-S reversals prior to removal of the optical background signal. b) Noisein the reversal correlated amplitude is dramatically reduced with the removal of the probebeam background.
can be effectively removed in measurements and provides a dramatic improvement in
the long term stability of the the measured gyroscope response (Fig. 4.9). The back-
ground removal subtracts residual beam walk and Faraday rotation effects as well as
any other change such as photoelastic effects on the windows to the inner vacuum
space.
The high Bz measurements remove most of the background, but retain sensitivity
to the Bx and Lx fields (Eq. 3.24). These fields have been shown to drift over longer
timescales than a typical reversal interval. Therefore, high Bz measurements provide
both the probe background and a small constant contribution from the comagnetome-
ter. A 0.01 µgauss shift in the Bx field would provide an estimated 6 fT response
from the comagnetometer in the high Bz limit. Since only a monitor of the back-
ground offset between each reversal is of interest, this contribution is for the most part
neglected, but remains an important technical detail in background measurements.
108
4.2 Comagnetometer Zeroing
The long list of second and third order effects in Section 3.2 lead to possible systematic
effects. These can influence the signal to obscure a measurement of anomalous fields.
A scheme to stabilize these parameters to a small fixed value has been developed in
Ref. [28]. These procedures are referred to as “zeroing routines” despite only ever
reaching a small value in the parameter that is being “zeroed”. The following section
describes the implementation of the zeroing routines in the CPT-II comagnetometer
with special interest given to any correlated changes with reversals of the apparatus.
4.2.1 Quasi-Steady State Measurements
The lists of contributions to the signal in Section 3.2 all describe the steady state
response of the comagnetometer. The fast damping of the nuclear spins (Eq. 2.72)
implies that the steady state response can be achieved in a few seconds, so a quasi-
steady state measurement scheme has been developed [28]. A modified square wave
can be applied
B(t) = B0 tanh(s sin(ωt)) (4.8)
where ω is the frequency of measurement and s describes the sharpness of the tran-
sition. This function has the feature that all time derivatives are continuous and the
field efficiently transitions between field values in finite time. With a typical sharp-
ness value of 10 and ω = 2π × 0.2 Hz, the field transitions in 600 ms (Fig. 4.10). As
long as ω ωn and the slope B0ωs |Bnωn|, the 3He spins will efficiently move
from one steady state value to another. The comagnetometer response can be aver-
aged and compared at the constant values of the waveform. An elaborate LabView
program provides the modulations and selects the appropriate intervals to serve as a
quasi-steady state lock-in amplifier.
109
Time (s)
Am
plitu
de
(arb
. units)
Squares=10
0 1 2 3 4 5 6 7 8 9 10−1.5
−1.0
−0.5
0.0
0.5
1.0
1.5
Figure 4.10: A example square wave modulation at ω = 2π × 0.2 Hz. A sharpness value of10 transitions the field in 600 ms.
4.2.2 Compensation Point
Several hours of optical pumping produces an equilibrium polarization of 3He at a
modest applied field of tens of mgauss. Once the equilibrium polarization has been
reached, the field can be slowly and smoothly lowered to find the compensation point.
Any sharp changes in the applied magnetic field will induce transverse oscillations of
the spins that decay with the long T2 of the 3He.
There are two common methods used to identify when the compensation point has
been reached. At large fields (Baz 20 µgauss), DS dominates the signal (Eq. 3.20).
As the applied field nears the compensation field, Bz in the denominator no longer
suppresses terms in numerator. A sharp dispersive response in the signal is observed
due to the product of ByBz in the numerator. The sharpest part of this feature is
close enough to the compensation point that an automated routine can provide fine
adjustments to zero the Bz field (Section 4.2.3).
Near the compensation point, the nuclear spins are quickly damped (Eq. 2.72).
Far from the compensation point, a By field in the form of a modified square wave
excites the nuclear spins. By comparing the damping time as the Baz field is slowly
adjusted as before, a dramatic decrease in the damping time can be observed. At this
110
point, the automated zeroing routine can adjust the applied field to precisely locate
the compensation point.
4.2.3 Zero Bz
Assuming that the comagnetometer is already close to the compensation point as
described in Section 4.2.2, a simple quasi-steady state modulation procedure can
evaluate the value of the applied compensation field. Considering Eq. 3.43, the re-
sponse of a By modulation is proportional to Bz. In effect, the modulation takes a
derivative of SL with respect to By,
∂SL∂By
∝ Bz/Bc. (4.9)
By comparing the response to a By modulation at various Bz, a zero crossing can be
identified. A numerical simulation of SL for a By modulation at various Bz is shown
in Fig. 4.11. Near the compensation point, a simple two step measurement can
determine the slope and the crossing point. Common parameters for this procedure
and other zeroing routines is in Table 4.2.
It is not clear that this procedure produces a Bz field that is precisely zero. There
is a finite contribution from other higher order terms (Eq. 3.19), the largest of which
is −Ωz/(γnBc). Considering the rotation of the Earth, this would cause a shift in the
Bz value of 230 fT which is much smaller than the resolution of the magnetic field
coil. Noise from the computer contributes an RMS variation of 2 pT to the Bz field.
Likewise, drifts in the 3He polarization lead to a typical variation of 20 pT in Baz
over the course of an hour, with longer term drifts following the average of the 3He
polarization. By adding up the contribution to the current from the low noise current
supply and the computer, a value for Bc can be determined better than 0.5%.
111
(a) Bz > 0
∆S < 0
∆S = 0 ∆S > 0
(b) Bz = 0 (c) Bz < 0
Sig
nal
(ar
b.)
By
Time (s)0 5 10 15
Figure 4.11: The difference in response to S switches sign on either side of the zero crossing.Image Credit: [28].
4.2.4 Zero By
Once a small value for Bz has been reached, a procedure to reduce the effects of a
drifting By field can be performed. This is similar to the zeroing of the Bz field except
the roles of the fields are reversed,
∂SL∂Bz
∝ By + βnyBc
+ LxγeRtot
+γeΩx
Rtotγn(4.10)
where here the higher order terms amount to measurable offsets. In terms of stabi-
lizing the signal, it is most pertinent to reduce the effects of a drifting Bz field since
this value changes with the 3He polarization.
The presence of βny in Eq. 4.10 does not play an appreciable role in a Lorentz
violation search. The value of βny is already constrained to less than 1 fT [29] and its
contribution to the signal is suppressed by Bz/Bc which is a small number. The noise
on the By coil from the computer is 4 ngauss and the zeroing variations are typically
50 ngauss.
112
4.2.5 Zero Bx
If the same technique as described for By and Bz is used to zero Bx, a zero crossing
will be avoided. However, a second derivative in Bz will provide the appropriate
response,
∂2SL∂B2
z
∝ γeRtot
(Bx
Bc
− LyγeRtot
− γeΩy
Rtotγn
). (4.11)
If Bz is already close to zero, an asymmetric modulation of Bz will act as a second
derivative. Reference to Section 3.2 indicates that there are a large number of terms
proportional to B2z , so it is likely that Bx is not actually zero. However, Bx provides a
value such that a B2z modulation is suppressed. Since Bz is the most quickly drifting
parameter, this is a useful procedure.
The Bax field has similar noise characteristics to Ba
y , and Eq. 4.11 has many more
anomalous field contributions (Section 3.2.9). These are already constrained below
the resolution of the magnetic field coil. Measured variations in the Bax field are
25 ngauss.
4.2.6 Orientation Dependent Zeroing Shifts
These zeroing routines have been developed previously, but we have suggestively
displayed the largest higher order terms. It is clear from Eqs. 4.10 and 4.11 that
an orientation dependence in the zeroing routines will appear with reorientation of
CPT-II. Figure 4.12 shows a clear out-of-phase behavior for each of these fields. More
noise persists on the Bax value from a B2
z modulation (Eq. 4.11) whereas the Bay value
comes from a Bz modulation (Eq. 4.10). The overall offset comes from contributions
of beam misalignment2, remnant magnetic fields, and lightshifts. Remnant magnetic
fields are expected on the order of 20 µgauss and are difficult to evaluate without an
2This likely results from pumping along the x- and y-directions (Section 3.2)
113
−540 −450 −360 −270 −180 −90 0 90 180 270 360 450 540 630 720 810 900−25.0
−24.5
−24.0
−23.5
−23.0
−22.5
−22.0
Heading (degrees)
Applied
Fie
ld for
Zer
oing
(µga
uss
)
Bx
By
Figure 4.12: Baz : Reverse. Shifts of both the Bx and By zeroing routines with position.
Fits to this behavior provides 0.36 µgauss for the By amplitude and 0.38 µgauss for the Bxamplitude. Both responses are 90 out of phase as expected.
additional sensitive magnetometer. In practice, offsets larger than 100 µgauss imply
that the lasers are poorly aligned.
In terms of the applied field,
Bay = Bc
(− Lx
γeRtot
− γeΩx
Rtotγn
)−Boffset
y (4.12)
where the offset Boffsety is explicitly considered. As predicted, the Ba
y field is more
negative when the comagnetometer is in the east position. Similar behavior holds for
Bax,
Bax = Bc
(+ Ly
γeRtot
+γeΩy
Rtotγn
)−Boffset
x (4.13)
which agrees with Fig. 4.12 in predicting a more positive Bax field in the north position.
The orientation dependence of the zeroing routines can lead to a systematic bias in
a reversal measurement if the fields are not re-zeroed in every position. In specifically
considering the rotation terms with the product of estimated small offsets in the Bz
value, this orientation dependance contributes a < 1 fT correlation with reversals
114
Zero Bz Zero By Zero Bx Zero Lx
Modulation TypeSymmetric Symmetric Antisymmetric Antisymmetric
By Bz Bz Bz
Amplitude (µgauss) 3.9 3.3 9.9 79Frequency (Hz) 0.2 0.4 0.4 0.4Step Size (µgauss) 3.3 5.2 5.2 (0.05 V)Lock-in τ (ms) 30 3 3 3
Table 4.2: Typical zeroing parameters for CPT-II.
along the E-W axis from an offset in the By field. The effect along the N-S axis is
further suppressed by an extra factor of Bzγe/Rtot to < 0.01 fT. These effects are
smaller than those required in this thesis and have been ignored in measurements.
The phase of the lock-in amplifier is arbitrary, but consistent throughout all of
the measurements in this work. In the “Reverse” orientation, the slope of the Bx, By,
and Bz zeroing is positive. The By and Bz slopes become negative in the “Forward”
orientation, but Bx remains positive since it is determined by B2z . This implies that a
positive voltage applied to the Bx coil produces a field oriented the opposite direction
from the probe beam propagation as shown in Fig. 3.9. This is in agreement with the
shift in Fig. 4.12.
4.2.7 Zero Lx and ~sm
The probe laser is nominally linearly polarized. A small circular polarization leads to
a finite Lx. Likewise, this circular contribution will lead to a finite pumping rate. If
the circular component to the probe polarization can be removed, then both of these
effects have been reduced. The circular component of the probe beam can be adjusted
using a stress plate to squeeze a glass microscope slide and introduce a birefringence
to compensate for birefringence in the beam path. Coarse adjustment of the stress
plate is controlled by screws on the mechanical mount with finer control provided by
a voltage controlled piezo stack wedged between the glass and the mount.
115
To evaluate the birefringence in the probe beam, the pump beam is blocked to
depolarize the atoms. Depolarized atoms reorient with the local magnetic field and
the pumping of the probe beam. Near zero field, a clear optical rotation response
follows the pumping direction from the probe beam. This response can be observed
by adjusting the Baz field by -79 µgauss (Table 4.2) which is approximately the value
of Be. The sign reversal with pumping orientation provides clear signal for a feedback
mechanism.
When the pump beam is blocked by the shutter, the 3He polarization starts to
drift, and a quickly changing Bz field dominates the signal response. Shuttering the
beam for less than 0.5 s is sufficient to locate the zero crossing without domination
by the drift. This procedure is too fast to be performed by hand without a significant
loss of 3He polarization and must be automated for accurate adjustment of the stress
plate.
This procedure does not truly zero the Lx contribution during normal comagne-
tometer operation, but instead removes the average contribution over the entire probe
beam. Any inhomogeneities in the pump and probe overlap will lead to an offset. An
alternative is to use the LxLz term in Eq. 3.43 to remove the contribution that par-
ticipates in measurements, but this often presents difficult to interpret results. The
method described here is sufficient for these experiments. Even if there remains a
finite Lx contribution, it is valuable to maintain Lx at an unknown but stable offset.
4.2.8 Zero Ly and Lz
The Ly term provides a first order contribution to the signal; however, there is no
source of light along the y-direction to produce a lightshift. One might think that
a misalignment of the pump laser can project along the y-axis to introduce a non-
negligible Ly. Such a shift was directly introduced in an attempt to develop a Lz
feedback (Section 4.4.2) by injecting a small fraction of the pump beam along the
116
y-axis. A scan of the pump wavelength across the resonance should produce a clear
dispersive shape from the lightshift. This behavior is not observed and is more con-
sistent with the pump and probe beams determining the axes of the comagnetometer.
Therefore, sensitivity to Ly is ignored.
The Lz term is controlled by adjusting the pump beam wavelength. In the presence
of a single optical transition in K, it would be simple to adjust the pump beam
wavelength to the resonance of the D1 transition in K. The 3He buffer gas creates
a poorly measured pressure shift, so several absorption profiles have been measured
to determine the central wavelength of the D1 transition in the comagnetometer
cell. Precise determination of the central wavelength of these profiles are limited by
multiple systematic effects including interference fringes and spontaneous emission
from the laser.
The lightshift from the nearby D2 transition shifts the zero lightshift point from
the center of the D1 absorption line (Eq. 2.10). We predict a shift to a higher
wavelength of +0.0044 nm from the D2 line and estimate the zero lightshift point to
be at 770.084 nm as measured on our Burleigh WA1500 wavemeter3. This wavelength
is in good agreement with effects that depend on the value of Lz (Section 4.4). The
wavelength is continuously monitored on the Burleigh wavemeter through an optical
fiber vacuum feedthrough and manually corrected for deviations larger than 2 pm.
4.2.9 Zero Pump-Probe Orthogonality
CPT-I requires an elaborate feedback system to maintain the pump and probe beam
orthogonality over several hours (Eq. 3.17). The thermal isolation of the bell jar in
CPT-II reduces such mechanical drifts. The zeroing procedure for the orthogonality
described in Ref. [28] cannot be implemented using a single DFB diode laser; it
3An HP Agilent 86120B wavemeter gives a reading of 770.088 nm for the same light source, so itis clear that there is no absolute calibration.
117
does not provide enough pumping power to evaluate the contribution at high pump
intensity under normal operating conditions to evaluate α.
There are three degrees of freedom that contribute to “good” pump and probe
alignment: the pump beam, the probe beam, and the Baz field. Proper alignment is
achieved when the pump and probe beams are orthogonal with the pump beam and
Baz field also parallel. This is a challenging parameter space to explore. We have
found success in limiting the search space by placing a flat mirror on the quarter
waveplate to retroreflect the pump beam. It is simple to evaluate the retroreflection
on the back of the aperture (Fig. 3.11). If the pump beam is well centered on the
window and the retroreflection centered on the aperture, its alignment is complete.
Further adjustments are made to the probe alignment. It is roughly centered on
the two vacuum windows that are 30 cm apart. The vertical alignment is evaluated by
blocking and unblocking the pump beam. Overlap in the pump and probe produces a
clear optical rotation signal that can be reduced through adjustments to the vertical
alignment of the probe beam. These can typically be reduced to below 0.5 mrad of
optical rotation.
A final check is evaluated by raising the Baz field after executing the Bx zeroing
routine. Fine adjustments to the pump beam alignment can reduce the change in op-
tical rotation to < 5 mrad. If the pump beam is adjusted, adjustments to the probe
beam vertical alignment are repeated. The entire alignment procedure is performed
prior to pumping out the bell jar and left in place for the duration of the experiment.
If there is a change in the mechanical equilibrium from evacuation of the bell jar, these
changes are only small effects and are deemed acceptable. A scheme that automat-
ically feeds back on the atom response would be preferable, however the procedure
described here has led to consistent results and low drifts in the comagnetometer
signal.
118
4.2.10 Zeroing Convergence
Each of the described zeroing procedures depends on the values of the fields already
being small. Therefore, it is insufficient to consider each routine independently. Close
to the compensation point, these procedures converge on typical values after several
iterations. It is important to zero the Bz field frequently as this is the most quickly
drifting parameter due to its dependance on the 3He polarization. We commonly
observe coupling between the By and Bx routines due in part to the linear dependance
of the Bz field on By and the quadratic dependance on Bx. Similarly, the Lx and By
contributions are also coupled. Zeroing over several iterations cycles through these
couplings and converges on stable values. Values that only require small adjustments
from repeating the zeroing routines after a 20-30 minute interval are considered stable
if the Baz field has also tracked the 3He polarization.
The Lx zeroing routine causes a significant drop in the equilibrium 3He polarization
from briefly blocking the pump beam. This procedure requires additional repetitions
of the Bz zeroing to ensure that the compensation point has been properly readjusted
before other routines are executed. Adjustments to the Lx routine typically lead to
small refinements to the Bay field in the By zeroing procedure. The particular schemes
executed are discussed in Section 5.1.2.
4.3 Calibration
The gyroscope contribution provides a large systematic offset in the reversal correlated
amplitude along the N-S direction. A discrepancy between the sensitivity of the
comagnetometer and the calibration leads to noise in repeated measurements. For
stable amplitude measurements (< 1 fT), the calibration technique should have a
precision of at least 1/300 and be accurate to within a few percent. New calibration
119
methods have been developed in this work to identify systematic effects and improve
calibration precision and accuracy. These techniques are discussed here.
4.3.1 Traditional Calibration: ByBz Slope
The leading prefactor P ez γe/Rtot in Eq. 3.43 determines the sensitivity of the comag-
netometer to anomalous fields. From an experimental perspective, it is simple to
define the calibration as the appropriate value by which to multiply the voltage to
determine the response in fT,
β = κSL (4.14)
where β is the field to be measured. This includes amplification by both the lock-in
amplifier and the photodiode amplifier which have remained fixed throughout the
experiment.
In CPT-I, an anomalous field is difficult to apply directly. A calibration method
based on the zeroing of the Bz field has become the “traditional calibration” technique
[28]. For Bz . Rtot/γe, the ByBz dependance is well described by
S(ByBz) =P ez γeRtot
R2tot + γ2
e (Bz + Lz)2
(ByBz
Bc
)(4.15)
where only the largest term in the denominator has been considered. This form
presents a clear dispersion curve in Bz and can be observed under a symmetric By
modulation (Fig. 4.13). The slope at the zero crossing can be used to determine κ in
terms of known parameters
κ =( Bc
∆By
d
dBz
(∆S))−1
(4.16)
where ∆S is the change in signal from the modulation amplitude ∆By. κ can easily
be extracted from the two point zeroing that occurs every ∼ 200 s due to drifts in the
120
−125 −75 −25 25 75 125−0.010
−0.005
0.000
0.005
0.010
Bz (μgauss)
MapTwo Point Zeroing
ΔS/Δ
By (
V)
Figure 4.13: Baz : Forward. Map of the symmetric By modulation as a function of Bz.
The zero crossing is used to “zero” the Bz field. Near this point, sampling two points canefficiently determine the crossing.
polarization. The step size in Table 4.2 determines the separation of the two point
zeroing in Bz.
The “traditional calibration” has been validated in CPT-I against a known rota-
tion. A piezo stack wedged between the optical table and an immobile block gently
rocks the table at frequencies below 1 Hz. Monitoring the six degrees of freedom of
the optical table on inductive non-contact position sensors and comparing the motion
to the signal response gives agreement to within 10%.
4.3.2 Ωy Rotations
The ability to measure the Earth’s rotation rate with a signal to noise of 103 in CPT-II
reveals a clear 20% discrepancy in a rotation calibration compared to the traditional
method. The value of κ is consistently smaller for the rotation measurement. The
leading sensitivity to Ωy is (Eq. 3.31)
S(Ωy) =P ez γeRtot
Ωy
γn
(1− γnQ(P e)
γe
)(4.17)
121
0 90 180 270 360 450 540 630 720−2.0
−1.5
−1.0
−0.5
0
0.5
1.0
1.5
2.0
Heading (degrees)
Rev
ersa
l A
mplitu
de
(mV
)
Figure 4.14: Baz : Forward. Amplitude response of the comagnetometer under 8 reversals
of the platform verses different headings. The dominant effect here is the sensitivity to Ωy
from Earth’s rotation. An offset and linear drift in the signal are removed through stringanalysis (Section E.3). The phase of the response is within 3 of the N-S axis in the laband within the absolute orientation uncertainty.
where the second term gives a valuable 0.5% correction to the response. In Prince-
ton, NJ, the predicted amplitude response for a reversal measurement4 is S(Ω⊕y ) =
271.2 fT. There are a number of background effects (Section 4.1) and additional ro-
tation terms (Section 4.4) that have already been accounted for .
Figure 4.14 determines a calibration of 142.0 pT/V which disagrees with the
188.0 pT/V found using the traditional method (Section 4.3.3). The phase of the
response is within 3 of the expected N-S axis in the lab.
Rotations enter the Bloch equations in the same way as anomalous fields, so one
would more likely trust the rotation method to calibrate sensitivity to an anomalous
field. For the Lorentz violation experiment, this method of calibration is insufficient
since a sidereal variation is searched for within the Earth’s rotation rate measurement.
An independent method of calibration is required to separate these effects. In the
following sections, we describe several measurements to investigate this discrepancy
and develop an easily repeatable and robust calibration method.
4Here, we have made the reasonable assumption that Q(P e) ≈ 5.
122
4.3.3 Ωx Sensitivity Through Zeroing
An orientation dependance in the zeroing of the Bx and By fields has already been
discussed in Section 4.2.6. The rotation terms that cause these shifts can be used as a
check in the calibration value. For example, an Ωx term contributes to the zeroing of
the By field (Fig. 4.15). One would expect rotational dependencies to be small, but
the γe/Rtot prefactor enhances the effect to produce a shift amplitude of 0.420 µgauss.
Equation 4.12 predicts the amplitude of the expected shift in ∆Bay . Earth’s rota-
tion
Ω⊕xγn
= −∆Bay
Rtot
γeBc
(4.18)
can be predicted in magnetic units with known values of Rtot and Bc. In general, Rtot
is not directly measured and is instead inferred from the width in the Bz resonance
in Eq. 4.15 (Fig. 4.13). A least squares fit to Fig. 4.13 provides Rtot = 362 1/s and
Lz=0.201 µgauss. These values predict a Earth’s rotation rate to be 355 fT which is
in poor agreement with the expected value near 270 fT.
The Rtot value is in agreement with the traditional calibration technique. It is
likely that a magnetic field gradient broadens the resonance in Fig. 4.13 to increase
the measured Rtot value (Section 4.3.6). All of the atoms experience the same rotation,
so a rotation produces no gradient. If a gradient has broadened the resonance, the
correct value of Rtot from Eq. 4.18 should be 276 1/s.
Measurements of the Bx and By zeroing shifts have traditionally not used the
reversal procedure and string analysis described in Section 4.1.2. For the measurement
in Fig. 4.15, the Bz and By field are zeroed in each location as the the platform
incrementally moves 7.5 to the next position. These measurements are susceptible
to linear drifts in the Bx and By fields that are likely from thermal drifts. It is unclear
what causes the kink at 300, but reversals and the string analysis would produce a
cleaner measurement.
123
−270 −180 −90 0 90 180 270 360 450 540 6305.6
5.8
6.0
6.2
6.4
6.6
Heading (degrees)
Applied
Fie
ld for
By Z
eroi
ng
(µga
uss
)
Figure 4.15: Baz : Forward: Ba
y for the By zeroing as the heading of the comagnetometeris reoriented. The phase of the total response deviates by 5 from the expected alignment.Note the sign reversal of the shift from Fig. 4.12 for flipping the spin orientation.
4.3.4 Ωz Rotations
Under typical conditions, rotations about the z-axis are normally suppressed by 10−5
compared to rotations about the y-axis. The CPT-II implementation introduces the
ability to apply a continuous ∼ 0.1 Hz rotation about the z-axis, several orders of
magnitude larger than Earth’s rotation rate. Because the comagnetometer is much
more sensitive to rotations about the y-axis as opposed to the other directions, any
applied rotation about the x- or z-axis should consider a projection along the y-axis
to determine the net effect. Inspection of Eq. 3.33 reveals several terms that can be
varied to determine the Ωz dependance of the signal. Nominally, Lx < 0.1 µgauss
and is difficult to control with the stress plate. The product ByΩz provides a simple
method of determining the coefficient of an Ωz term by rotating the platform at a
constant speed and varying the By field.
Figure 4.16 displays the signal response under rotations of the apparatus for nine
values of the of the By field detuned from the nominal Bay field determined in the
south position. The apparatus rotates 1080 rotation at 0.1915 Hz determined by the
124
0 50 100 150 200 250 300 350 400−0.1
0
0.1
0.2
0.3
0.4
Time (s)
Sig
nal
Res
pon
se (
V)
Figure 4.16: Baz : Forward. Response of comagnetometer to several 1080 revolutions of the
platform at 0.1915 Hz at different By offsets. The net shift is modified by detuning the Bay
field. While rotating, there is a variation in the response periodic with a full rotation dueto small variations in tilt.
programmed velocity profile of the servomotor5. There is a periodic signal associated
with each rotation interval. Tilt sensors indicate a correlated response from wobble
while the platform is in motion. The tilt has been optimized to produce small changes
for positions 180 apart while the experiment is at rest. A simple digital filter removes
this variation before averaging the shifted signal. The shift is negative for positive
detunings of the applied By field and positive for negative detunings. The nominal
zeroing value is repeated in the first, fifth, and ninth rotation to check against possible
mechanical drifts from continuously rotating the platform.
The response in Fig. 4.16 can be repeated for several rotation frequencies6
(Fig. 4.17). The linear behavior in the shift with adjustments in By is as predicted,
5There is a smoothing function applied to the velocity profile, but this rate is accurate to 1%.6A 0.3 Hz rotation is within the capability of the motor, but moving the 1500 lb apparatus
this fast is an impressive sight and is not commonly performed for fear of damaging the rotationmechanism.
125
−12 −10 −8 −6 −4 −2 0 2 4 6 8 10 12 −0.2
−0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
By Detuning (μgauss)
Shift
Duri
ng
Rot
atio
n (
V)
Ωz(1)= 0.0957 Hz
Ωz(2)= 0.1915 Hz
Ωz(3)= 0.2872 Hz
Ωz(4)=-0.0957 Hz
Figure 4.17: Baz : Forward. Shift in comagnetometer signal under several rotation frequen-
cies of the platform. A misalignment in the comagnetometer axes and the rotation axismixes the sensitivity to Ωy and Ωz. The change with By detuning can distinguish the twoeffects.
as is the sign change in reversing the direction of Ωz. The decreasing shift for
increasing By detunings for positive rotations is also in agreement7 with Eq. 3.33.
The slopes of the fitted lines can provide a sensitivity measurement (Table 4.3).
In general, these values disagree with the traditional method by 30%, but are within
a few percent of the Ω⊕y measurement. These measurements indicate that rotation
measurements are consistent but do not agree with an equivalence to magnetic fields.
It is interesting that the Ω⊕y and Ωz measurements differ by a few percent. There
exist higher order ByΩz terms (Eq. C.27), but the leading −ByΩ2z/(γnBc)
2 term only
contributes a 0.1% change to the measured slopes. It is more likely that there is
additional flexing of of the breadboard while the experiment undergoes a uniform
centripetal acceleration. Such an effect could explain to a linear shift in the measured
slope as observed in Table 4.3.
This effect could be compensated for using the tilt sensors; however, the AOSI
tilt sensors have a bandwidth of 1 Hz and the acceleration from a rotation pushes
7Note that the ByBz slope is negative in the “Forward” coil orientation (Fig. 4.13).
126
Method Sensitivity Uncertainty Discrepancy
Ω(1)z -147.8 pT/V ± 0.6 pT/V 1.27
Ω(2)z -149.4 pT/V ± 0.6 pT/V 1.26
Ω(3)z -150.1 pT/V ± 0.8 pT/V 1.25
Ω(4)z -144.5 pT/V ± 1.2 pT/V 1.30
ByBz -188.5 pT/V - 1Ω⊕y -143.05 pT/V ± 0.26 pT/V 1.32
Table 4.3: Comparison of sensitivity slopes for the Ωz measurements in Fig. 4.17. Theuncertainty represents the statistical uncertainty from the fit. In general, the rotationmeasurements disagree with the traditional method by 30%.
them into the nonlinear region of their calibration curve. We also expect that a large
rotation can drive the comagnetometer away from the compensation point (Eq. 3.19),
Ωz/γn ≈ 5 nT which is a reasonable fraction of the compensation field of 244 nT in
these measurements. The Baz field remains at the same value determined by the Bz
zeroing while the apparatus is at rest for all measurements (Section 4.2). None of
these effects are large enough to account for the discrepancy with the traditional
calibration method.
Horizontal Intercept
The shift is not symmetric about the nominal Bay zeroing value. The horizontal inter-
cept in Fig. 4.17 can be used to estimate the alignment of the platform’s rotation axis
with the axes of the comagnetometer. The comagnetometer is most sensitive to rota-
tions about the y-axis, so we consider the case where the y-axis of the comagnetometer
is tipped into the rotation axis of the platform by a small angle θ,
S(Ω) =P ez γeRtot
(− ΩBy
γnBc
cos θ +Ω
γnsin θ +XΩ cos θ
)(4.19)
127
where only the leading rotation terms are considered and X contains the terms relat-
ing to Lx. Simplification of this expression for θ 1 yields
∆S =P ez γeRtot
( Ω
γn
)(− ∆By
Bc
+ θ +X ′)
(4.20)
where we explicitly express the signal shifts ∆S and magnetic field detuning ∆By as
displayed in Fig. 4.17. It is clear from this expression that both θ and X ′ contribute
to the horizontal intercept. Estimates of the probe lightshift give Lx ≈ 100 fT, so X ′
is on the order of 10−7. The intercept at 11 µgauss implies θ ≈ −0.26. The probe
beam has been aligned by centering the beam to within 1 mm on each of the vacuum
tubes windows along the 30 cm optical path. The precision of the alignment is on
the order of the measured value.
The measurements rely on the sensitivity to rotation to dominate over other offsets
as indicated by the convergence of the slope at different rotation rates to a similar
intercept. The rotation dependence of the zeroing (Section 4.3.3) sets the nominal
zeroing value to within a 1 µgauss offset which can contribute to an estimate of θ.
Here, By was zeroed in the south position, so the offset does not contribute. It would
be possible to adjust the stress plate and look for a shift in the horizontal intercept
to confirm the estimate of Lx. Products of rotations like ΩxΩz (Section C.1.5) may
play a role in features in the signal during a continuous rotation, but these have been
filtered along with the tilt effects.
4.3.5 Low Frequency Bx Modulation
The steady state solution does not provide any more obvious options for developing a
meaningful calibration technique. However, the low frequency oscillatory response of
the comagnetometer reveals new insight (Eq. 2.81). A low frequency Bx = B0 cos (ωt)
128
0.1 1 10
0.01
0.1
1.0
Frequency (Hz)
Out-
of-P
has
e R
esponse
(V
)
Figure 4.18: Baz : Reverse. Out-of-phase response of the signal to a low frequency modulation
of the Bx field. The modulation amplitude is 1.30 µgauss. The line represents a least squaresfit to points below 1 Hz.
modulation provides the following response
S(B0 cos(ωt)x) = −Pez γeRtot
(ωB0 sin(ωt)
γnBc
)(4.21)
with the replacement Bc ≈ −Bn to be consistent with the experimentally known
values. All of these values are readily available and can provide a simple determination
of P ez γe/Rtot.
The low frequency response demonstrates clear linear suppression of the mod-
ulation as a function of frequency as long as ω ωn (Fig. 4.18). The measured
amplitudes are positive which is in agreement with the coil in the “Reverse” orienta-
tion, and the sign of the response is also in agreement with the orientation of the Bx
coil. This technique was developed late in the Lorentz violation search and detailed
measurements of all of the calibration techniques under the same conditions were not
available for direct comparison.
With Bc = 179.9 nT and B0 = 1.30 µgauss, the measured sensitivity is 129 pT/V.
At the same time, the traditional calibration provides a value of 155 pT/V. This
129
20% difference is similar to the discrepancy between values from the traditional cali-
bration and rotations. This strongly suggests that a systematic effect influences the
traditional calibration method.
An automated calibration method for using the a slow modulation of the Bx
modulation has been developed. A 0.645 µgauss cosine modulation over two full
periods is applied to the Bx coil to return a sine wave response. There is a jump
in the signal from quickly shifting the Bx coil to the start of the cosine waveform8.
The second period is recorded to reject any discontinuous features at the start of the
waveform. The response is measured at 0.5 Hz then at 0.3 Hz. The slope between
these two points and the known value of Bc from the Bz zeroing determines the
sensitivity.
The drawback of the low frequency Bx calibration is that it is inherently slow.
The procedure typically takes 12 s to execute. In principle, a measurement at one
frequency can determine the slope if the line is assumed to pass through the origin.
The raw values from each measurement are saved individually along with the slope
in case a comparison with a one point zeroing is of interest. It has been sufficient to
use the two point zeroing.
The product of the low frequency Bx calibration and the gyroscope response pro-
vides an effective measurement of Earth’s rotation of 277 fT. This is within 2% of
the predicted value of 271 fT. The dominant systematic effect is likely in the mea-
sured value of Bc. This value is determined from the sum of the applied magnetic
field to reach the compensation point. There is already an expected 2 nT remnant
magnetic field inside the shields. It would be possible to confirm this experimentally
by reversing the directions of the spins and measuring the compensation field in each
orientation. This is a moderately time consuming process and was not checked ex-
perimentally over the course of measurements. Of the calibration methods described
8A sharper discontinuity is present from applying a sine waveform from the discontinues deriva-tive.
130
in this section, the low frequency Bx modulation is the most efficient and accurate
method.
4.3.6 Traditional Calibration Revisited
The slope at Bz = 0 in Fig. 4.15 determines κ and can be decreased by an inho-
mogeneity in the Bz field. A dBz/dz gradient broadens this resonance and increases
the measured value of κ. In the presence of a linear gradient, the Bz zeroing will
correctly identify the center of the gradient as the zero. The net effect can be simply
represented as the sum of two resonances at different Bz values spatially separated
by the width of the probe beam,
∂S
∂By
=1
2Bc
[ P ez γeRtot(Bz + δBz)
R2tot + γ2
e (Bz + δBz + Lz)2+
P ez γeRtot(Bz − δBz)
R2tot + γ2
e (Bz − δBz + Lz)2
](4.22)
where δBz represents the difference in magnetic fields of the two resonances. The
gradient
dBz
dz=
2δBz
d(4.23)
where d is the width of a uniform probe beam. Figure 4.19 shows a simulation
of the resonance broadening due to a 10 µgauss/cm gradient with Rtot = 276 1/s.
Equation 4.15 fits reasonably well to the broadened resonance with Rtot = 307 1/s. It
is entirely reasonable that this effect was not observed previously. The astute reader
may have noticed that Rtot 6= Rp + Rm + Rense + Re
sd in Table 3.2 which is likely due
to the broadening of the gradient.
The gradient not only influences the slope of the Bz resonance, but it also affects
the sensitivity of the comagnetometer. The product of the calibration and the gy-
roscope response clearly informs the effectiveness of the calibration techniques. By
applying a known dBz/dz gradient and zeroing the Bz field to remove any Bz offset,
the effective amplitude varies by 10% for a 5 µgauss/cm gradient using the traditional
131
−100 −50 0 50 100−150
−100
−50
0
50
100
150
Bz (μgauss)
∆S/∆
By (
arb. units)
No gradient10 μgauss/cm
Figure 4.19: Simulation of the Bz resonance with and without a dBz/dz = 10 µgauss/cmgradient. The net effect is to broaden the resonance and increase the value of Rtot thatnormally would have been extracted from a fit to Eq. 4.15.
calibration while the low frequency Bx calibration varies by less than 1% (Fig. 4.20).
Applying the gradient causes the 3He polarization to drift, so the case with no applied
gradient is repeated at the beginning and end of the measurement as a consistency
check.
The offset in Fig. 4.20a without an applied gradient indicates that a magnetic
field gradient already exists. The model in Section 3.2 assumes that the alkali-metal
polarization is a constant, but the pump propagation model in Eq. 2.15 indicates
otherwise. The model predicts that the polarization is roughly 50% at the center of
the cell and has a polarization gradient of 0.33/cm. This predicts a magnetization
gradient from the alkali of 21 µgauss/cm which can account for the measurement of
the gyroscope at 345 fT instead of 277 fT using the traditional calibration (Fig. 4.20).
Evaluating the slope of Eq. 4.22 to determine the effect of the gradient on the
traditional calibration yields
κ ∝ Rtot
P ez γe
(1 + δB2
z
γ2e
R2tot
)−1
(4.24)
132
320
340
360
380
µgauss/cm)
Cal
ibra
ted G
yro
scop
eA
mplitu
de
(fT
)
ByBz Calibration Under Applied Gradient
−5 −4 −3 −2 −1 0 1 2 3 4 5270
275
280
285
Applied dBz/dz Gradient (μgauss/cm)
Calibra
ted G
yro
scop
eA
mplitu
de
(fT
)
a)
b)
Figure 4.20: Calibrated gyroscope amplitude under an applied dBz/dz gradient. a) Tradi-tional zeroing technique. b) Low frequency Bx calibration. The traditional calibration isseverely affected by the gradient, while the low frequency Bx calibration is more robust.
with Lz = 0. As expected, the gradient will always broaden the resonance to increase
the reported value of κ. This effect was not reported in CPT-I, likely because the
alkali density was a factor of 4 lower at 160C.
4.4 Gyroscope Phase
The probe beam background effects discussed in Section 4.1 can lead to an apparent
shift in the phase of the comagnetometer gyroscope response of several degrees. Even
with these effects reduced below a significant level, phase shifts in the gyroscope
response persist from drifts in the vertical fields. These contributions lead to first
order sensitivity to reversals along the E-W direction.
133
4.4.1 Phase Behavior
The leading rotational sensitivity is in the y-direction
S(Ωy) =P ez γeRtot
Ωy
γn
(1− γnQ
γe− γ2
e
R2tot
(Bz + Lz)2). (4.25)
Rotations about the z-direction are normally suppressed by 10−5, but there is a finite
contribution to the rotations about the x-direction
S(Ωx) =P ez γe
RtotDS
Ωx
γn
(γe(Bz + Lz)
Rtot
)(4.26)
where again one of the leading correction terms is
φ =γeRtot
(Bz + Lz). (4.27)
Introduction of φ into the sensitivity to Ωy and Ωx terms provides
S(Ωx) =P ez γe
RtotDS
Ωx
γnφ (4.28)
S(Ωy) =P ez γeRtot
Ωy
γn
(1− γnQ
γe− φ2
)(4.29)
where this form is suggestive a phase shift with sin(φ) ≈ φ and cos(φ) ≈ 1 − φ2/2
for φ 1. It is only suggestive because Eq. 4.29 lacks the factor of 2 in the cosine
dependance of a small angle and expansion of DS in Eq. 4.28 will give a φ2 term.
Likewise, there is no correction to the Ωx sensitivity from the γnQ/γe term in Eq. 4.28
as there is in Eq. 4.29. This prediction holds for both the steady state solution
(Section 3.2) and the toy model (Appendix C).
To demonstrate this behavior, several maps of the gyroscope response were ac-
quired with the Baz field detuned from the compensation point (Fig. 4.21). Each point
represents 5 reversals of the apparatus. It is clear that a nonzero Bz introduces a
134
0 45 90 135 180 225 270 315 360−1.5
−1.0
−0.5
0.0
0.5
1.0
1.5
Heading (degrees)
Rev
ersa
l A
mplitu
de
(mV
)
Bz ~ 0
Bz =+13 μgauss
Bz = -13 μgauss
Figure 4.21: Baz : Reverse. Measured phase shifts from detuning the Ba
z field from thecompensation point. Each point represents 5 reversals of the apparatus and the lines are acosine fit with an arbitrary phase. This study was taken prior to the development of thelow frequency Bx calibration, so it suffices to forgo the calibration.
phase shift from sensitivity to Earth’s rotation along the x-direction of the comagne-
tometer. Prior to the detuning, the traditional calibration method varies less than
0.5% over the course of these measurements9, but does not determine the sensitiv-
ity10 with Bz 6' 0. Equations 4.28 and 4.29 imply that only a phase shift should be
observed to a large degree for γeBz/Rtot 1. Some of the applied Bz values are not
always small, so expansion of DS in Eq. 3.29 is not valid. In either case, a non-zero
Bz should decrease the contribution to Ωy.
The response can be quantified in terms of the amplitude and phase for several Bz
values . For negative detunings, the total measured amplitude increases (Fig. 4.22a).
There is also an asymmetry in the total amplitude with the sign of the Bz detuning.
This effect is likely due to the real change in sensitivity from the dBz/dz gradient.
The phase behaves with a linear change as expected with a slope of 1.4/µgauss
(Fig. 4.22b). Combined with the 200 ngauss fluctuations in finding the Bz zero value
9The measured calibration is 247 pT/V from the ByBz slope.10At the time of these measurements, only the traditional calibration was well-developed. Away
from the compensation point, this method does not return an accurate calibration.
135
−20 −10 0 10 201.1
1.15
1.2
1.25
Bz detuning (μgauss)
Tot
al A
mplitu
de
(mV
)
−20 −10 0 10 2040
60
80
100
120
Gyro
scop
e C
ross
ing
(deg
rees
)
Bz detuning (μgauss)
a) b)
Figure 4.22: a) Amplitude of the gyroscope response under a Bz detuning. b) Phase of thetotal gyroscope response from a Bz detuning.
(Section 4.2.3), E-W reversals are expected to be limited to ∼ 1 fT. This is slightly
smaller than the observed variation (Fig. 3.17).
Fluctuations and drifts in the lightshift from the pump laser contribute in an
identical manner from changes in the intensity or wavelength of the laser. It is simplest
to consider small changes in the pump wavelength to confirm the gyroscope response.
Changes in intensity dramatically alter the pumping rate and several hours must
pass for the 3He to reach a new equilibrium polarization. The traditional calibration
method can give a reasonable value of the sensitivity (Fig. 4.23a), but as with the
Bz detuning, the total amplitude behaves in some complicated way. As discussed in
Section 4.3, the gradient causes the shift in amplitude from the expected 277 fT.
The phase response behaves as predicted (Fig. 4.23b). Assuming all other system-
atic effects have been properly accounted for, the crossing has a heading of 90 when
λ = 770.087 nm. This is close to the zero lightshift point predicted in Section 4.2.8.
The slope provides a correction for pump wavelength drifts of 0.2/pm or equiva-
lently 1.07 fT/pm. The pump laser has demonstrated stability < 0.1 pm over the
short term, but can drift several pm/day. A lightshift feedback has been considered
to stabilize these long term drifts (Section 4.4.2).
136
769.95 770 770.05 770.1 770.15 770.2280
290
300
310
320
Pump Wavelength (nm)
Cal
ibra
ted A
mplitu
de
(fT
)
769.95 770 770.05 770.1 770.15 770.260
70
80
90
100
110
120
Pump Wavelength (nm)
Gyro
scop
e C
ross
ing
(deg
rees
)
a) b)
Figure 4.23: a) Amplitude of the gyroscope response calibrated using the traditionalmethod. b) Phase of the total gyroscope response from changes in the pump beam wave-length.
It is interesting that finite values of Bz and Lz serve to twist the sensitive direction
of the comagnetometer into the x-direction. In CPT-II, the vertically oriented spins
produce a signal when an anomalous field in the y-direction causes them to precess
into the x-direction. For spins already projecting onto the x-axis, a change in the
vertical fields causes the spins to precess in the horizontal plane to mix the x- and
y-axes.
An accurate measured phase of the gyroscope response is an important feature of
the comagnetometer for anomalous field searches. The gyroscope response determines
the axis of sensitivity to all fields including the interesting anomalous fields βey and
βny . Agreement between the axes as defined by the light fields and the rotation of the
Earth suggests that other fields are at most a small contribution to the total response.
At the compensation point and with the pump beam at the zero lightshift point, the
gyroscope crossing is measured at a heading of 88.5 in agreement with the absolute
certainty in the orientation (Appendix B).
137
4.4.2 Lightshift Feedback
The sensitivity of E-W reversals to drifts in the pump wavelength of 1.07 fT/pm in-
spired several attempts in producing a robust Lz feedback. Inevitably, these attempts
failed due to correlations between the pumping rate and lightshift. Results from these
attempts are briefly reviewed here.
Attempt 1: Traditional
A means of evaluating Lz has been developed in Ref. [28] using Eq. 4.15. A change
in the wavelength of the pump beam introduces an asymmetry in the Bz resonance
(Fig. 4.24a). Fitting these asymmetric dispersion curves to Eq. 4.15 provides a value
of Lz for each pump wavelength. These values determine a clean dispersion curve
as a function of wavelength11 (Fig. 4.24b). This curve predicts the zero lightshift
point to be at 770.0905 nm ± 2.5 pm which is close to the value predicted by other
means (Section 4.2.8). This value is likely to be dominated by systematic effects from
the dBz/dz gradient (Section 4.3.6). The full-width of the dispersion is 0.320 pm
(Eq. 2.8) and larger than that expected for the pressure broadened linewidth of K
in 3He. This is again consistent with the effect for a magnetic field gradient. The
traditional method is not a good choice for a Lz feedback since it lacks the desired 1 pm
precision in determining the zero lightshift point and requires many time consuming
Bz resonance maps.
Attempt 2: Injection of Lz into Ly
The comagnetometer has first order sensitivity to Ly (Eq. 3.43). It may be possible to
evaluate the Lz lightshift by injecting a fraction of the pump beam along the comag-
netometer y-axis. A shutter can block and unblock this lightshift beam to directly
measure the contribution from this source. As the pump wavelength is adjusted, a
11The errorbars here represent the statistical uncertainty only.
138
−150 −100 −50 0 50 100 150−2.0
−1.5
−1.0
−0.5
0.0
0.5
1.0
1.5
2.0
Bz (μgauss)
∆S/∆
By (
mV
)
769.9 770.0 770.1 770.2 770.3−15
−10
−5
0
5
10
15
Pump Wavelength (nm)
Fitte
d L
z (μ
gauss
)
b)a)
770.098 nm
770.018 nm
770.187 nm
Figure 4.24: a) Asymmetry to the Bz resonance from an Lz contribution at various pumpwavelengths. b) Map of the fitted Lz values from asymmetry in the Bz resonance. Thismethod identifies Lz = 0 at λ = 770.0905 nm ± 2.5 pm where the uncertainty accounts forthe statistical variability only.
769.5 770.0 770.5 771.0−60
−40
−20
0
20
40
60
Lightshift Beam Wavelength (nm)
Sig
nal
Shift
(mV
)
770.1 770.11 770.12 770.13 770.14770.090
770.095
770.100
770.105
770.110
770.115
770.120
Pump Wavelength (nm)
Zer
o C
ross
ing
(nm
)
769.0
a) b)
Figure 4.25: a) A third circularly polarized laser near 770.1 nm provides dispersive-likelightshift behavior when injected along the y-direction. This curve does not fit well to adispersion curve (red), likely due to unaccounted shifts from the associated pumping rate.b) The zero crossing of the lightshift beam (a) has a different value depending on thewavelength of the pump beam.
clear dispersive profile is expected. The additional pumping along the y-axis intro-
duces a shift in the By zeroing of ∼ 20 µgauss (Eq. 3.21) and can be compensated
for with the appropriate adjustment to the Bay field. These measurements have been
performed and were characterized by small signals and inconsistent results.
139
A change in wavelength of the pump beam changes both Lz and Ly simultaneously
as well as the pumping rate from both beams. These changes are directly coupled
and cannot be separated. Likewise, as a check, a change in the handedness of circular
polarization simultaneously reverses both the lightshift and the pumping rate effects.
An AOM was introduced in an attempt to break any coherence between the pump and
lightshift beams but provided little change. An independent circularly polarized laser
was later introduced as the lightshift beam and produced dispersive-like behavior as a
function its wavelength (Fig. 4.25a). This might at least provide an indication of the
zero lightshift point. A dispersion curve does not fit well indicating that the behavior
is more complicated than predicted. Also, the zero crossing shifts with a change
in pump wavelength (Fig. 4.25b). This scheme has been abandoned as a viable Lz
feedback.
Attempt 3: Large Lx from a Far Off Resonant Laser
The LxLz term in Eq. 3.43 provides an opportunity to determine Lz if Lx can be
varied. The linearly polarized probe beam already produces a small Lx. Adjustments
to the stress plate can increase the Lx contribution to the signal but also introduce an
additional background optical rotation signal. Wavelength adjustments of the probe
beam again lead to correlated pumping rate changes that enter to leading order in
the x-direction (Eq. 3.16).
Far from resonance, Rp ∼ 1/(ν− ν0)2 but |~L| ∼ 1/(ν− ν0). An available 200 mW
at diode at 808 nm would introduce a lightshift of 1.7 µgauss with a pumping rate
of only 4 1/s. To combine the two beams, a FF801-Di01 dichroic beamsplitter from
Semrock replaces the mirror before probe light enters the cell (Fig. 3.11) transmitting
808 nm light but reflecting the 770 nm light from the probe. A shutter once again
controls the 808 nm lightshift beam to compare correlated changes in the signal with
140
the introduction of the lightshift beam. An FF01-769/41 filter at normal incidence
prior to the photodiode prevents detection of the lightshift beam on the photodiode.
The lightshift beam along the x-direction produces shifts in both the Bx (Eq. 3.24)
and By (Eq. 3.21) zeroing. These are typically compensated with adjustments to Bax
and Bay on the order of 25 µgauss which can be dramatically changed by adjusting
the alignment of the lightshift beam. Repeated measurements of the zero lightshift
point have a scatter of 5 pm. Adjustments to the alignment of the lightshift beam
likewise shift the value by up to 10 pm. This scheme has been abandoned as a viable
Lz feedback due to large scatter and the sensitivity to the lightshift beam alignment.
Reference Cell
Ultimately, the ideas presented here have failed to produce a feedback mechanism
capable of stabilizing the pump laser to 1 pm. There has been discussion of including
an additional vapor cell on the platform. This would indeed provide a stable reference
with which to lock to the center of the absorption peak or the zero lightshift point
given any number of standard techniques. The additional hardware of heating a
second cell adds complexity to the apparatus that was not attempted in this work.
It has been sufficient to continuously monitor the pump wavelength using a Burleigh
WA-1500 wavemeter on board the platform.
4.5 AC Response
Traditionally, the steady state response is the main focus of the comagnetometer signal
[28, 43]. In developing the low frequency Bx calibration, a series of measurements
of the ac response of the comagnetometer have been performed. Some interesting
observations from these investigations are summarized here.
141
4.5.1 Low Frequency Response
The low frequency response is evaluated in Section 2.5.2 and used in developing an
accurate calibration (Section 4.3.5). The low frequency Bx modulation produces both
an in-phase and out-of-phase component not listed in Eq. 2.81. Expanding Eq. 2.78
to the next highest order in ω produces
P ex(t) ≈ P e
z γeRtot
(ωBx sin(ωt)
γnBn− (γeB
e/Q+ γnBn)ω2Bx cosωt
(γnBn)2Rtot/Q+O(ω3)
)(4.30)
correctly predicting the sign of both phase responses and the scaling with ω (Fig. 4.26).
However, the in-phase response fails to continue to follow the polynomial behavior in
ω at the lowest frequencies.
This behavior also persists for a low frequency By modulation. Further expansion
of Eq. 2.78 for a By modulation provides
P ex(t) ≈ −P
ez γeRtot
(ω2By cos (ωt)
(γnBn)2+
(2γeBe/Q+ γnB
n)ω3By sin (ωt)
(γnBn)3Rtot/Q+O(ω4)
)(4.31)
also correctly predicting the sign and frequency dependance of the phases (Fig. 4.26).
This analysis ignores the effects of Rntot which can become more significant at the
lowest frequencies. Keeping this contribution,
P⊥(t) = −1
2
[(Bx + iBy)P
ez γe(R
ntot + iω)/Qeiωt
γeλM eP ez (Rn
tot + iω)/Q+Rtot(γnλMnP nz + ω − iRn
tot)/Q+ ω(Rntot + iγnλMnP n
z + iω))
+(Bx + iBy)P
ez γe(R
ntot − iω)/Qe−iωt
γeλM eP ez (Rn
tot − iω)/Q+Rtot(γnλMnP nz − ω − iRn
tot)/Q− ω(Rntot + iγnλMnP n
z − iω))
](4.32)
142
10-4
10-3
10-2
10-1
100
Bx M
odula
tion
Res
pon
se (
V)
10-2
10-1
100
101
−101
−100
−10-1
−10-2
−10-3
−10-4
By M
odula
tion
Res
pon
se (
V)
Frequency (Hz)
Out-of-PhaseIn-Phase (negative)
Out-of-PhaseIn-Phase
a)
b)
Figure 4.26: Reverse: Low frequency response of the comagnetometer to modulations of thetransverse magnetic field (Bx = 1.3 µgauss, By = 3.9 µgauss). Deviations from the simplepolynomial behavior predicted by Eq. 4.30 and 4.31 are likely due to the dBz/dz gradient.A numerical simulation with a 20 µgauss/cm gradient in the Bz field correctly predicts thebehavior. The low frequency Bx calibration is used to determine the overall scale of themodel.
which reduces to Eq. 2.78 in the limit that Rntot → 0. Expanding this expression to
low orders of ω produces a frequency independent amplitude,
P⊥(t) = − (Bx + iBy)Pez γeR
ntot cos(ωt)
−iRtotRntot + γeλM eP e
zRntot + γnλMnP n
z
+O(ω) (4.33)
which contributes only an in-phase contribution to the Bx modulation and an out-
of-phase contribution to the By modulation. A numerical simulation indicates that
an extremely large Rntot ∼ 4/hour would be necessary to account for the behavior in
Fig. 4.26.
143
Section 4.3 indicates the importance of considering a magnetic field gradient.
Introduction of a 20 µgauss/cm gradient by the method in Section 4.3.6 produces the
curves in Fig. 4.26. Like the introduction of Rntot, the dBz/dz gradient introduces a
frequency independent amplitude12. The simulation incorporates Bn = 170 nT and
Be = 80 µgauss which are close to the experimentally measured values. The slope of
the out-of-phase Bx modulation determines the calibration for these measurements.
All of the behavior matches well except for the lowest frequencies of the out-of-phase
By response. These effects are suppressed by ω3 and it is suspected that the signal to
noise in this region is poor. The additional suppression of the By modulation forced
the applied field to be increased by a factor of 3 to achieve a reasonable signal to
noise.
Previously, the effects of a magnetic field gradient from the nonuniform K polariza-
tion has been considered in terms of contributing noise [28]. In this work, the effects of
the gradient have been measured in two ways, through the low frequency response and
the traditional calibration technique. Ultimately, the model is in agreement with the
experiment and demonstrates that the low frequency calibration method is insensitive
to the magnetic field gradient (Fig. 4.20b) as a feature of the K-3He comagnetometer.
4.5.2 AC Heater Coupling
In investigating the long term behavior of the comagnetometer, a correlation between
the ac heater amplitude and the signal was discovered. An analog feedback loop
adjusts the heater amplitude to maintain a consistent temperature of the oven. Fluc-
tuations in either the heater amplitude or its coupling strength to the signal cause
systematic drifts. Quickly turning on and off the ac heater generates a dramatic dc
shift in the signal.
12Simplification is complicated and not illuminating, so this result is not reproduced here.
144
40 50 60 70 80 90 100 110 120 1300
10
20
30
40
50
60
70
80
90
Heater Frequency (kHz)
Sig
nal
Shift
(mV
)
Figure 4.27: The ac heater shifts the comagnetometer signal. It is strange that it decreasesas 1/f3. Typical comagnetometer sensitivity calibrations at the time of this measurementare 400-500 pT/V.
This coupling is not through electronics or a ground loop. Blocking the pump
beam depolarizes the alkali atoms and no change in the signal is observed upon
turning on and off the heater; the atoms experience a field. The heater contains
resistive wire in a twisted pair to reduce the dipole moment of the wire. Likewise,
there is a blocking capacitor to remove any dc component to the ac signal. The
comagnetometer responds to a high frequency magnetic field as 1/f (Eq. 2.63). It is
interesting to observe that the coupling to the heater signal scales as 1/f 3 (Fig. 4.27).
The comagnetometer calibration at the time of this measurement was 400-500 pT/V,
so these shifts create an effective magnetic field of 20 pT. The coupling also scales
with the power to the oven and is slightly smaller for a triangle wave than a sine wave
which as unexpected.
If the spins precess quickly in an ac field, the time average of the length of the spin
projection vector decreases. Such a decrease would result in an offset in the zeroing of
magnetic fields (Section 4.2) that depends on the heater frequency. A comparison of
the heater at 49 kHz and 173 kHz does not provide a consistently observable shift in
the values found by the zeroing routines for the Bx, By, and Bz fields. This confirms
that there is no significant dc magnetic field introduced by the heater to shorten the
145
Heater Measured Field
Full Panel 900 pT/mALightshift Panel 900 pT/mA
Probe Window Entrance 1000 pT/mAPump Window 1000 pT/mA
Probe Window Exit 900 pT/mAStem 200 pT/mA
Assembled Oven 70 pT/mA
Table 4.4: Measured 50 kHz magnetic field pickup at the location of the cell from thesix heater elements on the oven. The stem is the smallest because it is farthest from thecell. It is interesting that the effects of the assembled oven reduce the coupling an order ofmagnitude.
effective spin vectors. Such an effect would scale as 1/f 2 and is inconsistent with
Fig. 4.27.
Despite the twisted wire pair, it is interesting to measure the magnetic field pickup
on a small coil for each of the six heating panels. Shielding the return wire on the
pickup loop is important to reduce capacitive pickup. At 50 kHz, a response of
∼ 1 nT/mA for each panel and 70 pT/mA at the center of the assembled oven (Ta-
ble 4.4). The stem heater produces the smallest effect because it is farthest from the
position of the cell. The effect at the center of the assembled oven is reduced by more
than an order of magnitude likely due to destructive interference from contributions
of all panels.
The effects of heater variations are reduced by operating the heater at the limits of
the amplifier bandwidth near 170 kHz to take advantage of the 1/f 3 behavior of the
coupling. Commercial high frequency amplifiers are bulky and do not fit conveniently
on the rotating platform. A home-built audio amplifier supplies 1 A of current at
170 kHz. The amplifier has been operated up to 300 kHz, but typical heater loads
of 200 Ω decrease its bandwidth. Operating at these higher frequencies reduces the
146
offset to 500 fT. During normal operation, the power to the oven varies 0.5-1% over
30 minutes and only contributes to long term drifts in the comagnetometer signal.
4.5.3 Sensitivity Optimization
Previously, a sensitivity optimization scheme has been developed to improve the
pumping of the alkali polarization which we refer to colloquially as “Routine 13”.
At 50% polarization, the value P ez γe/Rtot reaches a maximum. A high frequency
modulation of the By field at 110 Hz determines the prefactor in Eq. 2.78 when the
alkali atoms experience a low magnetic field. Adjustments to the alignment of the
pump beam as well as the wavelength provide an extremum in the response that
corresponds to the optimum polarization.
In the work on CPT-II, it has been noted that the extremum will often correspond
to a noticeably skewed alignment and low 3He polarization from poor spin-exchange
optical pumping efficiency. It is not clear why this is the case, but it is suspected
that the polarization gradient plays a significant role in this procedure. In this work,
it has been preferable to use the procedure described in Section 4.2.9 to align the
pump beam. Once aligned, adjustments to the intensity can optimize the sensitivity
by comparing the magnitude of the gyroscope response at various pump intensities.
4.6 Symptomatic/Enhanced 3He Relaxation
Comagnetometer operation benefits from an increased compensation frequency for
faster dynamics. A larger Bc also suppresses second order terms in (Eq. 3.43). Each
of these parameters are proportional to the 3He polarization. The polarization is not
limited by pump beam intensity, but rather by symptomatic relaxation effects of 3He.
Practical operation of the comagnetometer requires a compensation field of several
hundred nT to provide a compensation frequency of several Hz.
147
4.6.1 Self-Relaxation
The assumption that the magnetization of K and 3He is uniform within the comag-
netometer cell correctly captures most of the dynamics of the spins. However, the
asymmetry of the cell leads to a significant gradient in the dipolar field from Mn.
Strong self-relaxation of 3He along its own gradient is observed (Eq.2.24). This is the
dominant 3He relaxation mechanism in the comagnetometer and limits the equilib-
rium 3He polarization.
The largest contribution to self-relaxation is the asymmetry of the cell at the stem
where it has been pulled-off the vacuum manifold. To mitigate this effect, we prepare
the cell such that a blob of K metal plugs the neck of the cell. This is often achieved
by heating the cell with a heat gun to just above the melting point of K at 63C.
Surface tension holds most of the blob together. Gentle shaking and reorientation
of the cell while hot can maneuver the blob towards the stem. The blob typically
catches near the neck of the stem. After some practice, a sharp flick of one’s wrist can
snap the blob into position. As the cell cools, different 3He pressures push bubbles
through the blob until it hardens or surface tension holds it in place.
Typically, we evaluate the blob placement before reinstallation of the cell in the
shields. We place the cell in a large convection oven and uniformly heat the cell
until the K melts. If the blob hardens before the density of gas has equalized on
either side of the blob, we observe a dramatic drop of the blob down length of the
stem. This requires repositioning of the blob once again. If the cell passes this
test, we mount it inside the magnetic shields. Historically, the blob does not always
maintain its position due to the nonuniform distribution of heat around the oven. To
avoid frequent removals of the cell, CPT-II incorporates an adjustable stem heater to
redistribute heater power to the small volume behind the blob to pushing it back to
the neck of the cell.
148
0 10 20 30 40 50 60 70 80 90 1001.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2
Time (min)
Com
pen
sation
Fie
ld (
mga
uss
)
Figure 4.28: 3He relaxation in the dark in a field of 2.8 mgauss. While the decay mayappear to be a simple exponential, a fit disagrees (solid). The nonlinear piece of Eq. 4.34accounts for the nonlinear relaxation of the gradient of the dipolar field (dashed).
Optical access near the stem is restricted when the shields and inner vacuum space
are closed, so we evaluate the final placement of the blob by measuring 3He relaxation
in the absence of pumping (Section 2.2.2). Figure 4.28 shows a measurement of the
decrease Baz field to find the compensation point as the 3He relaxes in a fixed field of
2.8 mgauss. The solid curve represents a fit to a simple exponential decay. It is clear
that the decay decreases faster than the exponential as the polarization decreases. A
fit to a simple heuristic model
P (t) = P0e−at−bt2 (4.34)
more accurately describes the relaxation where P0 is the initial polarization and the a
and b coefficients contribute to the linear and nonlinear decay respectively. Successive
increases in power applied to the stem show diminishing values of b and lengthening
of T1. Once an equilibrium compensation field of 200-300 nT has been achieved, we
find this sufficient for practical comagnetometer operation. Occasional adjustments
of the stem heater can maintain the polarization for several weeks. After several
149
weeks of operation, the position of the blob shifts due to diffusion of 3He through the
liquid metal. One could wonder if the gradient in M e significantly contributes, but
in this case, placement of the blob would make little difference. It is clear that the
3He polarization is limited by the gradient in its own dipolar field.
4.6.2 Magnetized Comagnetometer Cell
After one particular re-installation of the comagnetometer vapor cell, it became dif-
ficult to locate the compensation point and virtually impossible to stabilize the po-
larization once found. Typical magnetizations of several mgauss were only achieved
at relatively high magnetic fields (Baz = 10 mgauss). Slowly lowering the Ba
z field
to find the compensation point as described in Section 4.2.2 caused the polarization
to quickly drop. Following the compensation point as it plummeted only served to
increase the relaxation effect. A new or severely increased field dependant polariza-
tion relaxation mechanism had been introduced. A measurement of the 3He relaxation
provided a clean exponential decay demonstrating that self-relaxation (Section. 4.6.1)
was not responsible. Our usual methods of measuring the compensation point failed
here, but a simple NMR tipping pulse and measurement of the initial free induction
decay (FID) amplitude provided a signal proportional to the 3He polarization.
While placement of the K blob was not the issue, a foreign object within the shields
can produce a fixed magnetic field gradient. Relaxation along gradients is suppressed
by (Baz )2 (Eq. 2.24) and could explain the sudden increase in relaxation upon nearing
the compensation point. An attempt to compensate for possible gradients by engaging
combinations of gradient coils failed. These fields work well for suppressing uniform
gradients but would could not possibly account gradients from a localized object.
Rather than open the shields, systematic measurements of the T1 of the cell provided
a surprising result: T1 ∝ Baz rather than the expected T1 ∝ (Ba
z )2 (Fig. 4.29).
150
−2 −1 0 1 2 3 40
20
40
60
80
100
120
140
Bz Field (mgauss)
T1 (
min
)
a
Figure 4.29: Measurements of T1 at various fields. Relaxation along a fixed gradient wouldgive T1 ∝ (Ba
z )2 whereas the measured response is linear.
This behavior has been previously measured in vapor cells exposed to large mag-
netic fields (∼ 10 kgauss). Figure 4.30b (reproduced from Ref. [84]) shows the mea-
sured T−11 in the low-field limit for two vapor cells, 14A and 20B, in which 20B has
been exposed to a large field while 14A has not13. The behavior of T−11 of the co-
magnetometer cell (Fig. 4.30a) is strikingly similar despite the orders of magnitude
difference in applied fields. The large increase in relaxation at low magnetic fields is
a signature of this effect. Note also the asymmetry in relaxation rates on either side
of applying zero field. Systematic studies the comagnetometer relaxation were not
undertaken as we are more motivated by tests of fundamental symmetries and this is
mostly a technical issue for restoring comagnetometer operation. The effect appears
to be caused by 3He relaxing at the walls of the cell after the walls have become mag-
netized in a large magnetic field. The authors suspect that ferromagnetic impurities
on the surface of the glass become magnetized at high field. The local magnetic field
gradient near these sites accelerate 3He relaxation. These effects are referred to as
“T1 hysteresis”, a change in T1 based on the history of magnetic fields applied to the
cell.
13The group at the University of Utah prefers to plot T−11 .
151
b)
−2 −1 0 1 2 3 410-1
100
101
Field (mgauss)
T1
(hou
rs-1)
−1
a)
Figure 4.30: a) T−11 as a function of applied magnetic field for the comagnetometer cell.
b) Low-field T1 hysteresis reproduced from Ref. [84]. Note the similarity in T−11 for the
magnetized cell (20B) with the magnetic field of the comagnetometer cell despite the ordersof magnitude separation in applied field.
Fortunately, magnetizing a comagnetometer cell does not render it useless. T1
hysteresis effects can be reversed by degaussing the cell [85]. Here, the cell is placed
between two poles of an electromagnet and rotated perpendicular to the field using a
dc motor as the field magnitude slowly decreases. This device is colloquially referred
to as the Electrified Cell Spinner (ECS) and is useful for systematic hysteresis studies.
Rotation of the cell in the decreasing field randomizes the ferromagnetic domains
within a site to reduce the net magnetization. We were able to approximate operation
of this device by placing the cell on a G10 rod within an 1.1 T electromagnet. The
cell was slowly spun by hand while a colleague ramped down the magnet over 60 s.
It seems important to continue to rotate the cell until it is fully removed from the
magnet in case a remnant magnetization exists even at zero current. Re-installation
of the cell after degaussing restored the cell to normal operation.
We did not intentionally magnetize our cell and how it occurred is still an open
question. We often carry the cell to a neighboring room to reposition the K blob. It
is possible that the cell was placed close to a set of magnetic coils providing a field of
at most ∼ 10 gauss while in transit. Considering that most cells do not suffer from
magnetization effects until exposed to 10 kgauss, this is a surprising occurrence. It
152
has been shown that even exposure to low fields (< 50 gauss) will cause the onset of
T1 hysterisis [85]. A factor of two change in T1 in a cell exposed to 30 gauss field has
been observed [84]. Also, work at the University of Utah reports that these effects
are largest for Pyrex cells. Our comagnetometer is constructed from higher density
aluminosilicate glass to prevent diffusion of 3He through the glass. Nevertheless, it
is clear that it is possible to magnetize the comagnetometer cell and care should be
taken to prevent these events.
153
Chapter 5
A Test of Lorentz and CPT
Violation
We performed a search for a Lorentz- and CPT-violating field coupling to the neutron
spin using the CPT-II apparatus described in Chapter 3. The reversal technique de-
scribed in Section 4.1 allows for removal of the long term drift in the signal response.
The careful understanding of the comagnetometer behavior in Chapter 4 provides a
clear understanding of the signal response. We search for sidereal oscillations in the
signal to constrain an anomalous spin coupling of an extra-solar origin and improve
the existing limit on the neutron spin by a factor of 30. These results have been pub-
lished in Ref. [30] and represent the highest energy resolution of any spin anisotropy
experiment. The following chapter provides additional details not included in the
publication.
5.1 Lorentz Violation Search
The Lorentz violation search includes data for 143 days from July 2009 to April 2010
(Fig. 5.1). Collection is separated into two distinct parts represented by the summer
and winter. The long time span provides an important separation between sidereal
154
250
260
270
280
290
300
N-S
Rev
ersa
ls (
fT)
3450 3500 3550 3600 3650 3700 3750 3800−30
−20
−10
0
10
20
30
E-W
Rev
ersa
ls (
fT)
Sidereal Day since January 1, 2000
Apri
l 201
0
July
2009
Pump Laser ReplacementComputer UpgradeImplement Bkg. Measurements
Reverse ReverseForward
Summer
Winter
Figure 5.1: Complete nine month collection interval for the Lorentz and CPT violationsearch. The large gap in the middle provided an opportunity for continued comagnetometerbehavior studies and improvements to the apparatus. The two large collection intervalsoccur during the summer and the winter. The sign of the calibration maintains a consistentcalibrated response despite a reversal of the spins.
and diurnal signals with the Earth on the opposite side of the sun while a point on the
surface faces the same star. Both the N-S and E-W signals are collected as they are
susceptible to different systematic effects (Section 3.4.3). Also, the direction of the Baz
field has been reversed as a further check of systematic effects. Early measurements
suffered from repeated interruptions evidenced by the shifts in the E-W alignment.
These issues were solved during the Fall of 2009 with several upgrades to the long term
operation of the apparatus. The winter data demonstrates noticeable improvement
in the robustness of measurements achieving uninterrupted operation over several
weeks.
5.1.1 Measurement Protocol
Long term measurements for the Lorentz violation search proceed as follows: The
apparatus typically starts in the south position with the magnetic fields and lightshifts
155
well-zeroed. Data acquisition begins, and a trigger signal is sent to the motor via the
Labview Datasocket server. The motor waits 7-10 s and reverses the experiment.
An odd number of reversals are performed every 22 s. An extra rotation of 90
occurs after the same delay. The signal at rest is averaged and the reversal correlated
amplitude determined by the string analysis.
Data acquisition pauses at the end of fixed number of intervals for the Bz zeroing
routine. This defines one record of data. The next reversal sequence repeats starting
in the west position. This pattern is repeated such that every other record makes
measurements along the E-W or N-S axis. Also, each time a different axis is measured,
the apparatus starts in the opposite position compared to the previous record along
that axis. This serves to average any effect related to the starting location.
Reversal measurements are separated into several large files ranging from 1.5-10
days in length. They are labeled by the starting sidereal day and span sd 3492.86
to sd 3756.33 (Table D.3). Short duration files typically indicate that an error inter-
rupted operation. Longer files are typically stopped when convenient to investigate
the most recent results of the experiment. A search for a sidereal variation relies on
several days of uninterrupted measurements. Over intervals less than one day, linear
drifts in the data fit well to a sine wave clearly distinguishable from a true sidereal
variation in longer measurements. Longer measurements beyond a week in length
begin to separate sidereal and diurnal effects.
Both computers save all time values in sidereal days from January 1, 2000, in
GMST which is convenient for later analysis of the sidereal signal. The calculation
from the local computer time is described in Appendix A. The two computers are
independently synchronized with a navy time server using freely available Dimension
4 software to query the server every 3 s and synchronize the computer clocks. Typical
corrections are on the order of 10 ms.
156
At the beginning of each record, the local sidereal time is determined. The times
during the record follow the hardware clock on the data acquisition card and are
calculated from the sample rate typically at 200 Hz. In parallel, the computer con-
trolling the motor saves the sidereal time at each instance that the motor begins to
rotate the platform. A comparison of times recorded for a simultaneous event shows
that the computers are well synchronized within 20 ms. This is sufficient to reliably
coordinate selection of 3-7 s samples of data separated by 22 s.
The computer onboard the rotating platform saves several classes of data referred
to as “fast”, “medium”, “relaxed”, and “slow”. These designations classify signals
into different types for ease in processing. The comagnetometer signal and other high
bandwidth parameters such as the tilt sensors and position detectors are sampled at
200 Hz and averaged to lower sample rates. The signal follows the sample rate in
Table D.2 while the other parameters are saved to the medium rate at 2.5-10 Hz.
Various temperature readings have long time constants and are saved at 1 Hz in
the relaxed data. The slow data acquisition handles coordination of slow bandwidth
measurements, namely the encoder and the wavemeter. These communicate through
USB devices and cannot be easily synchronized with the other data at the same high
bandwidth. These values are sampled every 3 s as determined by a software timing
rather than the hardware timing of the rest of the analog channels. Each of these
values are saved as single precision binary numbers except the sidereal time which
is saved as a double precision binary number. This sampling and averaging scheme
maintains manageable file sizes without significant loss in compression.
5.1.2 Zeroing Schedule
Frequent execution of the zeroing routines maintains sensitivity of the comagnetome-
ter to the anomalous spin-dependent fields and limits the effects of higher order terms
on the signal. The Baz field must be adjusted between 50-200 ngauss every 200-300 s
157
due to drifts in the 3He polarization. This timescale limits the length of consecu-
tive reversal measurements. Other fields only demonstrate significant drifts over the
timescale of hours. These timescales divide zeroing into the minor and major zeroing
routines. The minor zeroing routines occur before each record along the N-S or E-W
axis to maintain a small Bz field and evaluate the sensitivity of the comagnetometer.
The major zeroing routines occur every 7 hours1 and execute a sequence of all the
zeroing routines (Section 4.2).
The particular zeroing sequences executed in the experiment evolved over the
course of measurements. In the summer, the following sequences were executed
Minor Zeroing : Bz
Major Zeroing : Bz By Bz Lx Bz Bz By Bz Bx Bz.
This sequence constrained drifts in the Bz field to 50 ngauss between minor zero-
ing events, though it was typical to observe a 500 ngauss/hour decrease in Baz field
corresponding to loss of 3He polarization (Fig. 5.2). Closing the shutter for a few
seconds for the Lx zeroing drops Baz by 8 µgauss from loss of spin-exchange optical
pumping. Repeated zeroing of the Bz field compensates for the loss and tracks the
return to equilibrium. A polarization feedback routine similar to that described in
Ref. [28] was not implemented due to the steady decline in overall polarization. The
set point would likely need to be adjusted on daily intervals and could lead to daily
correlations.
A similar procedure continued into the winter data collection. At this point, op-
tical rotation background measurements of the probe beam were implemented (Sec-
tion 4.1.7). Periodically increasing the Bz field increases the 3He equilibrium magne-
tization by 25 µgauss by suppressing 3He relaxation along magnetic field gradients for
1This value is chosen such that it is not an integer multiple of a diurnal or sidereal day to avoida correlation at these frequencies.
158
3524 3525 3526 3527 3528 3529 35302.29
2.30
2.31
2.32
2.33
2.34
2.35
2.36
2.37
2.38
Sidereal Day
3H
e M
agnet
izat
ion (
mgau
ss)
Figure 5.2: Summer: Adjustments to the Baz field using the Bz zeroing to track the 3He
magnetization. The regular drops in magnetization occur from closing the shutter duringthe Lx zeroing.
a significant fraction of the record (Eq. 2.24). This leads to additional drifts in the Bz
field, so the Bz zeroing executes twice in each minor zeroing routine to compensate.
In file 3692.34, a procedure to maintain the 3He polarization during the zeroing
was added. A routine to maintain a high Bz field (HBz ) raises the Baz field by the
same 0.65 µgauss as the background measurements for 8 s. An additional routine
for a longer high Bz field (HBz L) raises the Baz field for 60 s to compensate for the
polarization loss from the Lx zeroing. The zeroing sequence becomes
Minor Zeroing : Bz HBz Bz HBz Bz HBz Bz
Major Zeroing : Bz HBz Bz HBz Bz By HBz Lx HBz L Bz HBz By Bz HBz Bx Bz.
In file 3734.60, the low frequency Bx calibration was introduced to improve upon the
traditional method in the Bz zeroing routine (Section 4.3.5). At this time, the range
on the stress plate piezo was insufficient to fully compensate for the birefringence
of the probe beam. Rather than open the bell jar to adjust the mechanical screws
159
for increased range and delay the experiment, collection continued with the following
refinements to the routines
Minor Zeroing : Bz HBz Bz HBz Cal Bz
Major Zeroing : Bz HBz Bz HBz Bz By HBz Bx Bz HBz Bz
where the Cal method performs the low frequency Bx calibration and the Lx routine
has been removed. This provided reasonable performance for the remainder of the
Lorentz violation search.
The major zeroing routines execute on a timer every 7 hours along with the minor
zeroing routines at the end of a reversal sequence. The orientation dependance of the
Bx and By zeroing routines (Section 4.2.6) introduces possible shifts in the Bx and By
fields depending on the orientation of the apparatus on this timescale. Begining in file
3498.72, execution of the major zeroing routines was delayed until the apparatus was
aligned along the N-S axis. This is accomplished by comparing the encoder reading to
be within 2 of the starting heading modulo 180 reducing the offset in the By zeroing
but maximizing the effect in the Bx zeroing. In terms of drift, the offset from Bx is
suppressed by B2z , so it can be neglected as long as the Bz zeroing routine executes
regularly.
During the winter data collection, reading of the encoder values would fail after a
few days of operation from a firmware error in the JSB524 counter. The major zeroing
routines were executed manually at approximately 7 hour intervals for a short time.
Shortly after failure of the counter, accessing the DLL files would freeze the operating
system and require a hard restart of the computer. The counter at the base of the
experiment is difficult to access without some disassembly. Since the variations in
platform placement are too small to substantially influence the correlated reversal
signal (Section 5.2.5), a simple scheme of passing a text string containing one of the
160
four resting platform positions from the motor to the rotating computer through the
Datasocket server was introduced at file 3669.09. A simple comparison of the text
string allowed execution of the major zeroing routines with the experiment in the
correct position.
Just prior to file 3734.60, a clear orientation dependence in the Bx zeroing routine
was observed which had been previously obscured by noise in previous investigations.
At this point, major zeroing routines were executed only in the south position to avoid
long term shifts in both the By and Bx zeroing. Future experiments will likely use
a more sophisticated scheme of adjusting the Bay and Ba
x fields by a known amount
during each record to compensate for the Ωx and Ωy contributions to the Bz and
B2z terms respectively. These ideas have been discussed but not implemented at this
time.
5.1.3 Reliability
Many times during the acquisition, a failure would occur that would stop operation.
While much of the operation of the experiment is fully automated, several errors re-
quired human intervention. During the summer months, a NI-DAQ error would halt
operation from a miscommunication between the two data acquisition cards. This was
later solved with a computer upgrade with two separate PCI slots rather than one
slot and a PCI bridge. During the first few weeks of the winter dataset, the LabView
program on the onboard computer would lock up and fail to quit. This required a full
restart of the computer. It was later deemed to be a result of a faulty 24-bit counter
and a memory overwrite error. Also, in the early spring, Princeton’s Office of Informa-
tion Technology (OIT) would occasionally disable local internet claiming that illegal
devices were present on the network. In the short term, this was solved by connecting
increasingly longer ethernet cables to any available active ethernet port and having
many stern conversations with OIT representatives. An active internet connection
161
allows for reliable synchronization of computer clocks with the navy timeserver, and
active network access is required to reliably pass NI datasocket triggers.
Both computers contain minimal software for undistracted operation. Windows
updates have been disabled to prevent unwanted interruptions for the data acquisi-
tion system. Several interlocks disable the ac heater if certain components exceed
preset temperature limits. A simple error handling system determines when applied
values on the data acquisition system near the hardware limits. An email client sends
a daily email during normal operation as well as notifications of simple errors. The
email client can also send a text message to the operator’s cell phone. Uninterrupted
operation over three weeks has been achieved, though regular supervision is recom-
mended to gather long uninterrupted datasets.
5.1.4 Least Squares Fits
Due to the large systematic drifts in the CPT-I data, the amplitude of sidereal os-
cillation was determined by multiplying the comagnetometer signal by a sinusoidal
response of arbitrary phase and averaging the product over integer periods (Sec-
tion 3.3.2). The elimination of this long term drift allows for a more direct analysis
by directly fitting the gyroscope signal to a sidereal oscillation with arbitrary phase
using a least squares fit
SR = Ax cos(Ω⊕t) + Ay sin(Ω⊕t) + c (5.1)
where SR is the reversal correlated amplitude, Ax and Ay are the amplitudes of the
two sidereal phases, and c is a constant offset. This fitting occurs independently for
both the N-S and E-W measurements. Nominally, the E-W measurements would
have zero offset, but imperfect alignment with the gyroscope signal creates a small
but finite offset.
162
272
274
276
278
280
282
N-S
Rev
ersa
ls (f
T)
3670 3672 3674 3676 3678 3680 3682 3684 3686 3688 3690
10
15
20
25
Sidereal Day from Jan 1, 2000
E-W
Rev
ersa
ls (f
T)
Figure 5.3: Example long term collection of the N-S and E-W amplitude with a sinusoidalfit to a sidereal frequency.
The sidereal time is saved in GMST, but the lab is located at -74.652 W lat-
itude. Prior to fitting, the times are adjusted by t → t − 74.652/360 to LST for
straightforward interpretation of the AX and AY amplitudes in celestial galactic co-
ordinates (Appendix A). Fits occur on datasets that are continuous and longer than
one day. Fits to several sidereal cycles are preferred for greater separation of sidereal
and diurnal effects.
5.1.5 Combined Measurements
The N-S results are most sensitive to variations in comagnetometer sensitivity and
calibration while the E-W results are sensitive to variations in orientation of the
comagnetometer and drifts in the Bz and Lz fields. Throughout a given day, mea-
surements of both of these directions distinguishes variations of these two systematic
effects. Furthermore, a true sidereal variation provides an out-of-phase response in
163
both signals, distinct from any Earth-frame field. The amplitudes of independent fits
in the N-S and E-W measurements (Eq. 5.1) can be combined
AX →−(ANSX / sinχ,AEW
Y
)AY →−
(ANSY / sinχ,−AEW
X
)(5.2)
where χ = 40.3453 is the latitude in Princeton, NJ. At t = 0, the celestial x points
overhead which corresponds to a projection south for an experiment in the Northern
Hemisphere. A sidereal component is most sensitive when oriented in the equatorial
plane of the Earth, therefore, the sensitivity of N-S measurements to anomalous fields
is suppressed by the the sine of the latitude.
5.2 Correlation Analysis
The least squares method can be limited by long term drifts in the signal. These drifts
are smaller than those observed in the CPT-I collection but still persist. A correlation
analysis of the signal with each of the other recorded parameters has been performed.
The only finite correlations observed occur from drifts to the Lz and Bz fields. This
section will describe several parameters of importance and their contribution to the
analysis.
5.2.1 Calibration
The low frequency Bx modulation calibration scheme (Section 4.3.5) was not devel-
oped until file 3734.60 near the end of the Lorentz violation search. For consistency
in the measurement, the traditional calibration method has been used to calibrate all
files, though the total amplitude of the gyroscope response has been rescaled to 277 fT
as determined by the low frequency Bx calibration. In comparing the two calibration
164
3742 3744 3746 3748 3750 3752−129.0
−128.5
−128.0
−127.5
−127.0
−126.5
−126.0
−125.5
−125.0
−124.5
Cal
ibra
tion
(pT
/V)
Sidereal Day
Traditional
Bx(ω)
Figure 5.4: Long term comparison of the traditional and the low frequency Bx calibrationschemes. The traditional calibration has been rescaled by 20% to more closely match thevalue of the correct calibration. The traditional calibration is much more susceptible todrifts in the 3He polarization at the beginning of the file.
techniques, the traditional calibration is more sensitive to 3He polarization drifts at
the beginning of the file. Near sd 3746, both calibrations capture a sensitivity change,
but near sd 3749, the traditional calibration presents a drift that is not present in
the low frequency calibration. This is clearly a failure of the traditional calibration
as the rotation signal should be constant (Fig. 5.5). Overall, the low frequency Bx
calibration more closely matches changes in sensitivity of the comagnetometer.
5.2.2 Faraday Effect
Large square Helmholtz coils surround the experiment to reduce Earth’s magnetic
field and correlated optical rotation signals from reversal of the apparatus in this
field. No feedback scheme has been implemented on these coils, though a fluxgate
on the platform monitors changes in the local field along all three coordinate axes.
Magnetic field changes along the x-axis provide the largest contribution to the optical
rotation signal. A clear long term change in the field under reversals of the apparatus
(Fig. 5.6b) can be observed without a dramatic change in the corresponding signal
165
3742 3744 3746 3748 3750 3752
274
276
278
280N
-S R
ever
sals
(fT
)
3742 3744 3746 3748 3750 3752
274
276
278
280
N-S
Rev
ersa
ls (
fT)
Sidereal Day
a)
b)
Figure 5.5: N-S response calibrated with each of the practical techniques. a) Traditionalcalibration. b) Low Frequency Bx calibration. It is clear that the deviation at sd 3749 inFig. 5.4 is due to the traditional calibration and not a change in the sensitivity.
(Fig. 5.6a). Perhaps a correlation exists at sd 35005.5, but this is not consistent with
a correlation between sd 3499.5-3500.
These changes in local magnetic field are due to slow ramping (0.25-0.5 T/min)
of a 14 T superconducting magnetic in a nearby lab. Clear structure matching the
particular ramping routine rates can be observed under closer inspection of the flux-
gate response. The fluxgate is mounted 30 cm from from the rotation axis of the
platform and is also sensitive to the magnetic field gradient. Section 4.1.5 predicts
that the 2 mgauss change in field corresponds to 50 nrad of optical rotation. Given
the sensitivity of the comagnetometer, these changes are on the order of 0.5 fT and
below the typical scatter (Fig. 5.6a).
5.2.3 Pump Lightshift
Two diode lasers have served as the source of pump light with the replacement made
in the gap between the summer and winter data collection. The diode used in the
166
−24
−22
−20
−18
−16
−14
E-W
Rev
ersa
ls (
fT)
3498.5 3499 3499.5 3500 3500.5 3501
−4
−2
0
2
Exte
rnal B
-fie
ldA
mplitu
de
(mgau
ss)
Sidereal Day
a)
b)
Figure 5.6: a) E-W reversal amplitude. b) Long term correlation behavior of the magneticfield measured by the fluxgate outside the shields. The change in local magnetic field isfrom ramping of a 14 T superconducting magnet in a lab down the hall. These changes donot significantly influence the measurement.
summer was near the end of its useful lifetime and was subject to large drifts of
1 pm/day. Corrections for drifts in the pump wavelength can significantly improve the
stability of the measured reversal amplitude. The laser used in the winter collection
drifts 0.1 pm/day, so no correction is needed.
Significant drift in the pump beam wavelength introduces a change in the LzΩx
term (Section 4.4) and leads to a drift in the long term response of the E-W reversals
(Fig. 5.2.3a). It is clear that the drift in the pump wavelength is correlated with the
change in the E-W reversal response (Fig. 5.7b). Here, 30 consecutive wavelength
measurements are averaged and a cubic spline used to interpolate the values for the
E-W reversal domain. Prior to averaging, outliers have been automatically rejected by
considering groups of 10 points and rejecting any values beyond 2 standard deviations
from the mean.
167
−10
−5
0
5
E-W
Rev
ersa
ls (
fT)
770.089
770.090
770.091
770.092
Pum
p W
avel
egth
(nm
)
3548.5 3549 3549.5 3550 3550.5 3551 3551.5 3552 3552.5 3553−4
−2
0
2
4
6
E-W
Rev
ersa
ls (
fT)
Sidereal Day
a)
c)
b)
Figure 5.7: Pump lightshift drift corrections. The gap at sd 3552.5 is from a data acquisitionerror in the middle of the night. a) E-W reversals before correcting for changes in Lz. b)Long term pump wavelength measurements. c) E-W amplitude adjusted for drifts in thepump beam wavelength.
Measurements of the gyroscope phase shift described in Section 4.4 provide a
correction to the pump beam wavelength drift (Fig. 5.7c). This correction has been
applied prior to rescaling the reversal measurements to 277 fT and uses a value of
+1.33 fT/pm to account for the drifting pump beam2. This effect is second order in
the N-S reversals and need not be applied. The gap at sd 3552.5 is due to a data
acquisition error in the middle of the night. The experiment was promptly restarted
in the morning and continued without error until the end of the file.
2If the direction of the spins are reversed, the sign of this correction should also reverse.
168
0
2
4
6
E-W
Rev
ersa
ls (
fT)
3524 3525 3526 3527 3528 3529 3530−2
0
2
4
Sidereal Day
E-W
Rev
ersa
ls (
fT)
a)
b)
Figure 5.8: a) E-W reversal amplitude under the influence of Bz drift showing a clearasymmetry in the cental amplitude from starting in the east or west position. b) Reductionof the influence of the Bz drift using the m4 procedure and extrapolating to the time of theBz zeroing.
5.2.4 Bz Drifts
Like the pump wavelength drift, a drifting Bz field introduces sensitivity to Ωx. The
Bz field drifts on a faster timescale than the pump wavelength due to the chang-
ing 3He polarization (Fig. 5.2) and must be accounted for differently. E-W reversal
measurements show a clear bifurcation in the reversal measurements depending on
whether the experiment started in the east or west position (Fig. 5.8a). The Bz
field continues to drift the same direction, but asymmetry in the drift on individual
reversal measurements leads to a systematic offset with the starting position.
Early in the search, E-W reversals were averaged in pairs and adjusted using the
time derivative of the Bz field to account for this drift. More sophisticated procedures
described in Section 5.3.2 perform much better. The m2, m3, and m4 methods each
reduce the correlation of the reversal amplitude to Bz drift (Fig. 5.8b).
169
3542 3542.5 3543 3543.5 3544 3544.5
3542 3542.5 3543 3543.5 3544 3544.5
southnorth
eastwest
Enco
der
Rea
din
g M
odulo
90˚
Sidereal Day Sidereal Day
17.99
18.00
18.01
18.02
18.03
18.04
18.05
Figure 5.9: Typical encoder reading for positions of the platforms 180 apart. The positionsare taken modulo 90 for comparison on a fine scale. The encoder starts arbitrarily at the18 mark in the south position. The discontinuity shortly after sd 3542.50 comes frompausing the experiment and adjusting equipment on the rack.
5.2.5 Orientation
Inconsistent placement of the platform in the same position can introduce a systematic
effect. The position is recorded continuously every 3 s using an incremental encoder
with 0.001 precision. Figure 5.9 shows the encoder value at each rest location before
the next trigger. We expect that the scatter is due to imperfections in the teeth of
the gears. This variation is remarkably different in comparing the north and west
positions (Table 5.1). It is interesting to note that the points cluster along a lower
line and scatter upward asymmetrically. This is likely a result of the 0.2 backlash in
platform and overshoot. The servo motor comes to a particular value determined by
its own encoder, and the large momentum of the platform carries the platform beyond
this point. The servomotor encoder cannot recognize this overshoot or compensate
for it within the backlash.
Shortly after sd 3542.50, the wavemeter failed to record values over the GPIB. The
experiment was paused to remove the wavemeter from the lower shelf to reboot it.
170
Position Typical VariationNorth 0.001
South 0.004
West 0.012
East 0.006
Table 5.1: Typical variations in platform location as measured by the encoder. These aretoo small to observe in the reversal signal even in the E-W direction.
In the process, the equilibrium position of the experiment was inadvertently shifted
0.005 as seen by the discontinuity at the gap. For long runs, we observe a relative
drift of the platform and legs < 0.002 over 10 days. For scale, a 0.02 shift in the
position along the E-W direction would only produce a 0.1 fT change in the signal.
Fortunately, these effects are too small to contribute significantly to our measurements
and do not enter the analysis.
While the platform position is well-known from the encoder, the location of the
vibration isolation provides an additional free parameter. The mechanical springs and
rubber come to a new position and orientation with each reversal of the platform.
The relative position of the vibration isolation platform with respect to the aluminum
baseplate is recorded using two non-contact position sensors on either side of the
platform. This allows separation of a rotation from a translation in the recorded
signals. Each of the sensors is calibrated separately by shifting the platform and give
the same slope within 0.1%.
The reversal correlated orientation of the vibration isolation platform rises to an
equilibrium value over 10 hours. The scatter is less than 0.002 and the magnitude
of the drift is less than 0.010. As with the variations in the platform positioning,
these changes do not significantly influence the comagnetometer response at the 0.1 fT
level. The observed backlash from the vibration isolation platform is 0.09, which can
almost contribute at the 1 fT level, though all measurements are made with reversals
in the same direction.
171
−0.010
−0.005
0.000
0.005
0.010
Vib
ration
Iso
lation
Sta
ge O
rien
tation
3511.6 3511.8 3512 3512.2 3512.4 3512.6 3512.8 3513 3513.2 3513.4−0.015
−0.010
−0.005
0.000
0.005
Sidereal Day
E-W
(deg
rees
)N
-S (
deg
rees
)
Figure 5.10: Typical long term variations in the orientation of the vibration isolation stagerelative to a 180 reversal. Over 10 hours, the platform reaches an equilibrium orientation.
5.2.6 Long Term Signal Variations
The fits in Fig. 5.3 are limited by wiggles observed in the data. The source of these
variations is unknown. A correlation analysis with the signal and other parameters
such as temperatures, laser position detectors, water flow, etc. has been performed.
No direct correlation has been observed with any single parameter other than the
pump lightshift and Bz drift previously discussed. A fourier transform of the reversal
amplitudes may provide insight into time-dependant patterns in the response.
The longest sequence of uninterrupted data combines files 3669.09, 3677.13, and
3685.43. A fourier transform of this combined file produces the spectrum in Fig. 5.11.
An FFT requires equally spaced data, and the values in Fig. 5.3 have a typical spacing
of 480±6 s. A linear interpolating function creates a new domain with equally spaced
points of 100 s to meet this requirement3. The spectrum in Fig. 5.3 shows no peaks at
any of the low frequencies indicating that there is no regular pattern to the variations
3The interpolating function also accounts for minor computer slowdowns that are present in thedataset.
172
Frequency (Hz)
N-SE-W
Effec
tive
Magnet
ic √Fie
ldA
mplitu
de
(fT
/H
z)
103
10-4
10-5
10-6
101
102
Figure 5.11: Fourier transform of the longest uninterrupted measurements in Fig. 5.3. Thebin size is 1.8× 10−6 Hz. The E-W reversals are subject to Bz field drift and exhibit highernoise on short timescales. There is no peak at 4 × 10−5 Hz which corresponds with themajor zeroing routines.
in Fig. 5.3. Higher frequencies are limited by the averaging of the string points and
zeroing schedule. A peak at 1/480 Hz correctly indicates the spacing between the
original points as an artifact from the interpolating function (not shown). It is likely
that the larger noise in the E-W signal on short timescales is due to drifts in the Bz
field.
The N-S reversals in file 3519.53 present the largest variations all of the files with
excursions up to 1.5 fT from the central value (Fig. 5.12). The cause of these variations
is unknown, but this file provides a representative case to search for correlations with
other saved values. No strong correlation with any single channel has been observed.
Some trends arise and are only suggestive of a correlation as they do not persist.
5.3 Analysis Methods
The long collection time over nine months makes it difficult to vary experimental
parameters, repeat the experiment, and use the scatter in the result to determine
a systematic uncertainty. Instead, the same data is used repeatedly, but analyzed
173
3520 3520.5 3521 3521.5 3522 3522.5 3523 3523.5 3524 3524.5 3525274
275
276
277
278
279
Sidereal Day
N-S
Rev
ersa
ls (
fT)
Figure 5.12: N-S reversals in file 3519.56 demonstrate the largest drift and underlyingstructure in any of the measurements. No correlation with other saved values have beenobserved.
with differing processing techniques. The scatter in processing techniques provides
an estimate of the systematic uncertainty in the measurement. The following section
describes details relevant to data processing and determination of the systematic
uncertainty.
5.3.1 Final Processing Scheme
By the end of the collection, a global routine was developed to analyze all of the
data and incorporate all of the pertinent elements to the analysis. Data with the
experiment at rest is selected and averaged during each record. Prior to averaging,
a digital Butterworth low pass filter is applied to remove oscillations in the signal
from 3He precession between 5-10 Hz (Fig. 4.1). To prevent artifacts from the digital
filter, an extra 2 s prior to each sample is filtered and dropped prior to averaging to
the central value. The uncertainty is determined using the scatter and the ENBW
(Section E.2).
The string analysis determines the reversal correlated voltages and removes back-
ground drift (Section E.3). Five parallel processing techniques are added at this step
to account for higher order drifts and to determine the systematic uncertainty (Sec-
tion 5.3.2). The reversal correlated voltages are combined with the calibration and
174
a correction applied for drifts in the pump wavelength (Section 5.3.3). Outliers are
removed by manually selecting an appropriate domain in every file (Section 5.3.4).
These files are saved for straightforward analysis by the least squares fit with several
variations (Section 5.3.6). The results are combined to a final pair of amplitudes and
the systematic uncertainty determined (Section 5.4.1).
5.3.2 Processing Methods
Five different processing techniques have been developed to analyze each record. The
first is a straightforward application of the 3-point overlapping string analysis; the
next four modify this process. These techniques are referred to as “simple” and
“modification 1-4” using the shorthand s, m1, m2, m3, m4 for convenience. Likewise,
the intervals are analyzed with and without the removal of the probe background for
a direct comparison of the improvement.
Simple: s
For each element in a record, a three point overlapping string analysis is performed.
The overall sign convention is assigned according to the first measurement. The N-S
and E-W measurements are separated and each converted to the sign convention
north minus south (ns) and east minus west (ew).
The introduction of the probe background measurements in the winter introduces
secondary measurements. A 3-point overlapping string analysis on the background
measurements provides an overall correlation from the background which we refer
to as ns2 and ew2. Subtraction of the signal and background amplitudes demon-
strates improvement in the the short term scatter of reversal measurements. Further
improvement is achieved by considering the difference in the signal and background
with each reversal. The string analysis on these differences generates amplitudes re-
175
0 20 40 60 80 100 120 140 16032.2
32.3
32.4
32.5
32.6
32.7
32.8
Sig
nal
(m
V)
Time (s)
0 2 4 6
50
60
70
80
90
String Index
Str
ing
Poi
nts
(μV
)
a) b)
Figure 5.13: File 3669.09, record 1576. E-W reversals a) Example case of the disturbanceof the signal from rotating the apparatus and drifts in the Bz field. This first point hasno drift due to adequate time to settle during the zeroing routines. b) The resulting stringpoints from the record in (a). The change in drift causes the first point to be disjoint fromthe rest. The m3 and m4 processes compensate for the drift by fitting a line to the stringpoints. The m4 process disregards the first string point as shown.
ferred to as ns3 and ew3. In the case without a probe background measurement, ns2
and ew2 contain zeros, and ns3 and ew3 are identical to ns and ew.
Modification 1: m1
This modification considers an improvement to the determination of the uncertainty
of the signal at rest. In the case that the signal drifts over the averaging interval
(Fig. 5.13a), the standard deviation over-estimates the uncertainty. A least squares
fit to a line can remove the drift. These files follow the same structure as s and the
results are qualitatively the same.
Modification 2: m2
This modification reduces the effect of drift in the Bz field over the entire record
(Fig. 5.13a). The overall signal drifts to higher voltages from drifts in the polarization.
The drifting Bz adjusts the phase of the gyroscope signal (Eq. 4.28) and contributes
differently from the orientation of the apparatus; measurements 1, 3, 5, and 7 drift
176
less than measurements 2, 4, 6, and 8. Fitting the odd and even measurements to a
line with the same slope but opposite sign isolates this effect from the string analysis.
Due to the subtle changes in the probe background with each measurement, these fits
are performed only after removal of the background.
Modification 3: m3
Modification 2 provides a complicated method to remove the Bz drift from the signal
itself. This drift propagates into the string points, but an alternate approach in mod-
ification 3 with a linear fit to the string points can also remove this effect (Fig. 5.13b).
Two results are reported for the amplitude, the central value (ns3c/ew3c) and the
y-intercept (ns3z/ew3z). The center value is similar to the string analysis without the
fit. The y-intercept extrapolates back to the time of the zeroing and estimates the
value without Bz drift. Again, these results are only performed after direct removal
of the probe beam background.
Modification 4: m4
The most striking feature of Fig. 5.13b is the discrepancy of the trend of the first
string point with the rest. Comparison with the signal in Fig. 5.13a indicates that a
new trend in the drift occurs during the first reversal. The string analysis accounts
for a linear drift, not a change in the drift. It is likely that mechanical stresses on
the platform and settling of the vibration isolation stage cause the trend. So far,
observations of the vibration isolation stage on the inductive position sensors do not
account for this change, but subtle changes in the tilt below the resolution of the tilt
sensors could contribute to a rotation of Ωy.
The discontinuity in the string points is not present in every record, but consis-
tently dropping the leading point improves the scatter. We conclude that this method
best represents the reversal correlated amplitude. A consistent sign convention for
177
3450 3500 3550 3600 3650 3700 3750 3800
280
300
320
340
360
File
Tot
al G
yro
scop
e A
mplitu
de
Tra
ditio
nal
Cal
ibra
tion
(fT
)
Figure 5.14: Total gyroscope amplitude for each file in the Lorentz violation search using thetraditional calibration method. The influence of the magnetic field gradient is increased inthe winter collection and is smaller when the spins are reversed. The dashed line representsthe value from the low frequency Bx calibration.
the interval is maintained with the starting condition despite dropping the leading
point. Again, intervals are only considered after point by point removal of the probe
background.
5.3.3 Calibrated Data
A Matlab structure array organizes the various processed voltages for each file into
a single easily addressed structure. Each process is calibrated and corrected for drift
in the pump lightshift (Section 5.2.3). The low frequency Bx calibration was not
developed until file 3734.60, so for a consistent treatment of the data, the traditional
calibration is used throughout the files.
The traditional calibration provides an overall amplitude that is too large and
varies across the files (Fig. 5.14). The low frequency Bx calibration consistently
provides an overall amplitude of 277 fT and is insensitive to the magnetic field gradient
from the polarized alkali-metal vapor. The amplitude of the calibrated signal is
178
rescaled to 277 fT considering the contributions from the N-S and E-W data added
in quadrature4.
For the calibration, groups of 10 values are averaged. As with the pump wave-
length correction (Section 5.2.3), outliers are automatically rejected prior to averag-
ing. A cubic spline interpolates the calibration value for each amplitude to match the
original points. Two parallel blind analyses, each using different software (Matlab
and Genplot) evaluate the result in units of fT and check the consistency of the final
results.
5.3.4 Domains
Automated routines have been considered for rejecting outliers in the reversals similar
to those used for determining the calibration and pump wavelength. These and more
advanced routines are not perfect and have been rejected for removal of outliers from
the calibrated data. Every file is investigated by hand, and all obvious outliers are
removed. Typically, the first point in the file is shifted with respect to the rest of the
file much like the first string point in a sequence (Fig. 5.13b). These first points are
automatically removed and disregarded.
5.3.5 Averaging
Groups of consecutive points can be combined into a single value and uncertainty
prior to the least squares fit. A weighted average of groups of 1, 2, 6, 12, and 16/18
points are combined with a weighted uncertainty for each of the processes5. Not
every file has a number of points that easily divides by the size of the averaging
group. In this case, the remaining data is thrown away. Rather than remove the end
or beginning of a file, these regions are maintained. The dropped points are instead
4One could scale to the 271 fT as predicted by the theory, but we choose instead to rely on themeasurement and focus on setting a more conservative limit.
5An average of 16 or 18 is chosen to reduce the number of points ignored in the file.
179
evenly distributed throughout the file. In general, larger groups of averaged points
decrease the reduced χ2 of the least-squares fits to values slightly larger than 1. Care
has been taken not to average large groups such that less than 4 averaged points
remain per day. In this case, a sidereal frequency would be filtered from the data.
5.3.6 Least Squares Fit Refinements
Equation 5.1 describes the sidereal signal of interest in this work. A direct fit using
this method can be susceptible to other effects, so four additional methods have been
developed to fit the data. In the absence of systematic effects, these methods should
give similar results. Only a comparison of the results across all the methods can
determine the validity of the result.
Direct Fit
A direct fit uses a least squares fit using Eq. 5.1 to the N-S and E-W reversal measure-
ments separately. This is the simplest and most direct of the fitting techniques. This
technique returns amplitudes Ax and Ay and their associated uncertainties for each
file. Longer term files of several days demonstrate improved separation of possible
sidereal oscillations from variations in the background signal.
Background Fit
Drifts on a longer timescale than a sidereal day persist in the signal. A low order
polynomial can suppress the influence of these drifts on the fitted amplitudes without
removing a signal at a sidereal frequency. The order of the polynomial is selected
based on the length of the file. The order of the polynomial added to Eq. 5.1 is
limited by one less than the number of full sidereal days in the file. For example, a
180
file of length 3.15 sidereal days will be fitted to the following function
SR = Ax cos(Ω⊕t) + Ay sin(Ω⊕t) + c0 + c1t+ c2t2 (5.3)
where c0, c1, and c2 are the coefficients of the polynomial of order zero, one, and two
respectively. For these fits, a large integer around 3600 sd is subtracted from t, such
that the center of the polynomial is somewhere in the middle of the file. This leads
to a more efficient convergence of the overall fit6.
Single Fit
One can also consider the complete collection of each of the N-S and E-W measure-
ments as a single massive set (Fig. 5.1). The dc offset for each file differs particularly
for E-W reversals, so the average value of each file is removed before combining mea-
surements. This technique returns a single amplitude and uncertainty for each of the
total N-S and E-W collections. The single fit is subject to changing scatter across the
entire dataset but can fully distinguish sidereal and diurnal components in the time
domain.
Strong Fit
A true sidereal signal from a Lorentz-violating field will appear out-of-phase in each
of the N-S and E-W reversals. Both the N-S and E-W measurements from the same
file can be combined to constrain precisely the correct sidereal response between
orientations. The N-S and E-W reversals contribute to the measurement differently,
so they are both projected onto the equatorial plane for a direct comparison. The
6Figure 5.3 provides an example of the background fit technique where the polynomial backgroundhas been subtracted from both the data and the fit except for the c0 term for display purposes.
181
3676 3678 3680 3682 3684 3686 3688 3690−3
−2
−1
0
1
2
Local Sidereal Time
Cor
rela
ted S
ignal
A
long
Cel
estial
Axes
(fT
)
E-W Reversals N-S Reversals
Figure 5.15: Strong fit procedure on file 3685.43 where groups of 18 points have beenaveraged to a single value. The E-W reversals have been shifted forward in time by aninteger number of sidereal days so as not to overlap with the N-S reversals, and an extra0.75 sd matches the phases. The error bars on the N-S measurements are larger than on theE-W measurements due to the projections onto the equatorial plane. The time has beenadjusted to LST.
following transformation prepares a file for the strong fit:
REW (t)→ REW (t−N − 0.75) (5.4)
RNS(t)→ RNS(t)/ sin(χ) (5.5)
where R is the reversal correlated amplitude, N is an integer number of sidereal days,
χ is the latitude, and t is the local sidereal time. A direct fit is then applied to the
transformed dataset as in Fig. 5.15 where the mean value from each orientation is
again removed. In general, the N-S measurements will provide lower statistical weight
due to to the smaller spatial overlap with the equatorial plane. The results of this fit
on file 3685.43 are AX = −0.051 ± 0.093 fT and AY = 0.029 ± 0.093 fT where the
amplitudes are reported in terms of galactic celestial coordinates.
182
Single Day Fit
Each file can also be divided into single days and fitted by day. The advantage
of this technique is that it can provide over 100 amplitudes providing a reasonable
distribution of scatter in the measured amplitudes across the entire dataset as well
as avoiding bias from long term drift removal. Each file is divided into intervals of
at least one day where the first N − 1 days are equal, and the last one carries the
remainder. For example, a file of length 4.23 sd is broken into 3 days of 1.05 sd with
the last day of length 1.08 sd. In the case of a data acquisition gap greater than
0.1 sd, the gap is treated like the edge of a file. Each of these one-day intervals are
fitted to Eq. 5.1 and return an amplitude and uncertainty for both the N-S and E-W
measurements for each day selected.
5.3.7 Systematic Checks
Throughout the course of the experiment, three main parameters have been compared
to check against a systematic effect in the fitted amplitudes. (1) The N-S and E-W
measurements provide a separation of systematic effects in the gyroscope background
from Earth’s rotation. Furthermore, a Lorentz-violating field will appear out-of-phase
in each orientation to distinguish between Earth-based effects and celestial effects. (2)
The direction of the spins in the comagnetometer have been reversed twice throughout
the entire measurement. Equal duration of data with the spins up and down under
similar conditions have been collected during the winter months. This provides a
reversal of spin-based effects from the gyroscope and celestial fields uncorrelated with
other signals associated with operation of the apparatus. (3) The entire data collection
spans nine months with significant data collected in both the summer and the winter.
A comparison of the summer and winter data provides a complete separation of the
diurnal and sidereal phases with the Earth on the other side of the sun for the same
orientation in space-time.
183
5.4 Results
The large variety of processing techniques lead to a distribution in the measured
amplitudes which provides an estimate for the systematic uncertainty. Within this
distribution, several representative files are identified to determine a final measured
amplitude. The measured amplitudes are connected to parameters of the SME im-
proving the limit on bn⊥ by a factor of 30. The details of this analysis are described
here.
5.4.1 Representative File and Systematic Uncertainty
The fitting techniques, averaging, and processing methods lead to 136 amplitude mea-
surements for both phases of the N-S and E-W measurements where each amplitude is
a weighted average across the entire dataset7. A plot of the amplitudes shows remark-
able consistency (Fig. 5.16). The horizontal lines represent the weighted average of
the measured amplitudes, and only a few results are more than a standard deviation
from the mean. The m4 ns3c, ew3z file averaged in groups of 18 with the background
fit is well centered in three of the four distributions. The AEWY measurements present
a systematic shift correlated with alternating between the ew3c and ew3z methods.
The ew3z method accounts for known drifts in the Bz field, so a bias is no unexpected.
5.4.2 Sidereal Amplitudes
Given the representative file in Fig. 5.16, the individual amplitudes from each fit
can be displayed (Fig. 5.17). Here, the N-S and E-W measurements have been com-
bined and averaged as described in Section 5.1.5. These amplitudes represent a dra-
matic improvement over the equivalent results in the previous generation experiment
(Fig. 3.5).
7The strong fits are omitted in this comparison because they combine the N-S and E-W resultsinto a single amplitude pair.
184
−0.06
−0.04
−0.02
0
0.02
0.04
0.06
−0.1
−0.08
−0.06
−0.04
−0.02
0
0.02
Background Fit m4,ns3c m4,ew3z Avg 18
0 20 40 60 80 100 120 140−0.04
−0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0 20 40 60 80 100 120 140−0.05
−0.04
−0.03
−0.02
−0.01
0
0.01
0.02
0.03
0.04
0.05
AX (
fT)
NS
AY (
fT)
NS
AX (
fT)
EW
AY (
fT)
EW
Processing Index Processing Index
Figure 5.16: Measured amplitudes from the various processing, averaging, and fitting (ex-cept strong) techniques. The horizontal lines indicate the average of the amplitudes. Them4 ns3c ew3z file averaged in groups of 18 with the background fit is highlighted as a rep-resentative file. The scatter in the amplitudes is used to identify the systematic uncertaintyin this collection.
The summer data has a larger scatter of data points than indicated by their error
bars. We believe this is due to fluctuations of the background optical rotation. In the
winter, background measurements were implemented to compensate for these effects
and reduce the overall scatter. To obtain the final limit, the summer and winter data
are averaged separately, scaling the errors by their respective reduced χ2 before the
final average. A comparison of the results taken in the summer and the winter, with
the 3He spins up or down, and from the N-S or E-W amplitudes provides a check of
possible systematic effects (Section 5.3.7). Only the winter data is considered for a
comparison of the spin up and spin down states. The N-S error bar is larger than the
E-W error bar due to the reduced sensitivity to AX and AY in N-S measurements
185
−1
−0.5
0
0.5
AX (
fT)
3450 3500 3550 3600 3650 3700 3750 3800−0.4
−0.2
0
0.2
0.4
Sidereal Days from Jan 1, 2000
AY (
fT)
SU SD SMWN NSEW SU SD SMWN NSEW
−0.1
−0.05
0
0.05
0.1 AX AY
Sid
erea
lA
mplitu
de
(fT
)Spin UpSpin Down
Figure 5.17: Top Panel: Summary of the amplitudes AX and AY for each file as a functionof time. Upward and downward triangles indicate the direction of the 3He spins. Themean of the amplitudes is well centered on zero and the scatter improves as the experimentprogresses. Bottom Panel: Systematic comparisons: SU/SD - spin up vs. spin down (fromwinter data only), SM/WN - summer vs. winter data, NS/EW - amplitudes from N-S vs.E-W reversals only. The presence of a systematic effect will cause a shift from a reversal ofthe SU/SD, SM/WN, NS/EW parameters. Each pair agrees within its uncertainty.
(Section 5.1.5). A separation from a reversal of any of these parameters will indicate
a systematic effect, yet each provides a consistent result.
Artificial Lorentz Violation Simulation
It is possible that fitting an additional polynomial to the background can remove
sensitivity to a sidereal oscillation. As a check of both the processing techniques and
combining N-S with E-W measurements, a sidereal oscillation several standard devi-
ations larger than the limit is artificially added into the complete dataset. Execution
of the analysis yields the results in Fig. 5.18. These amplitudes agree within 1 aT
186
−0.5
0
0.5
AX (
fT)
3450 3500 3550 3600 3650 3700 3750 3800−0.6
−0.4
−0.2
0
Sidereal Days from Jan 1, 2000
AY (
fT)
Spin UpSpin Down
´´
Figure 5.18: Simulation: As a check of the fitting routines, a +8σ and −10σ siderealoscillation is introduced for AX and AY . The dashed green line represents the expectedcentral value of A′X = +0.168 fT and A′Y = −0.210 fT. These amplitudes are precisely thesame as those in Fig. 5.17 with the appropriate shift for the inserted sidereal signal.
to those presented in Fig. 5.17 accounting for the known shift of AX by +0.168 fT
and AY by −0.210 fT. This confirms that a sidereal oscillation can be detected and
also confirms correct treatment of the sign when combining N-S and E-W reversals.
Analysis of the subdivision parameters generates a similar effect for the systematic
reversals.
5.4.3 Final Average
In determining the final limit to report, six representative files have been identified
across the Genplot and Matlab investigations (Table 5.2). The final result is an
average of each of the six files. In each case, the summer and winter data have been
considered separately and the error bars scaled by the reduced χ2. This particular
consideration reduces the uncertainty by 10%. Files A and B represent the “best”
and “good” methods from the Genplot analysis. Likewise, files F and E are the “best”
and “good” methods from the Matlab analysis. C includes the first representative
187
AX AX Uncertainty AY AY UncertaintyFile Amp Total Stat Syst Amp Total Stat Syst
(fT) (fT) (fT) (fT) (fT) (fT) (fT) (fT)A 0.0020 0.0193 0.0183 0.0052 0.0314 0.0216 0.0197 0.0085B -0.0073 0.0231 0.0190 0.0128 0.0290 0.0252 0.0201 0.0138C 0.0077 0.0218 0.0188 0.0087 0.0362 0.0204 0.0186 0.0070D -0.0078 0.0224 0.0194 0.0087 0.0276 0.0206 0.0184 0.0074E 0.0031 0.0205 0.0182 0.0081 0.0303 0.0198 0.0182 0.0061F 0.0085 0.0195 0.0186 0.0043 0.0376 0.0199 0.0192 0.0050
Final0.001 0.021 0.019 0.010 0.032 0.021 0.019 0.010
Result
Table 5.2: Summary of the final amplitude measurements from the representative files. Thetotal uncertainty is the quadrature sum of the statistical and systematic contributions. Thefinal result is an average of each of the files.
file identified in Section 5.4.1, and D incorporates a “naive” analysis including all the
methods. Unless otherwise noted, the reported amplitude and statistical error are the
average of each processing method. The systematic error is the standard deviation of
the amplitude values.
The six files are as follows: (A) The “best” Genplot method considers the m4
ns3z and ew3z data with averaging of 1, 4, 10, 16, and 20 points. Both the single
day fits and background fits are considered. (B) A “good” Genplot method considers
all of the m3 and m4 processes and includes the ns3/ew3, ns3c/ew3c, and ns3z/ew3z
combinations. (C) The first representative file identified considers the m4 ns3c, ew3z
method with groups of 18 points averaged to a single value. The background fit
determines the amplitude and statistical uncertainty. The systematic uncertainty is
determined from the scatter in the amplitudes reported from the direct and strong
fitting techniques. (D) The “naive” method includes all processing, averaging, and
fitting methods described in the previous sections. (E) The “good” Matlab method
considers both the m3 and m4 processes and includes the ns3c/ew3c and ns3z/ew3z
combinations. Unlike the “good” Genplot method, the ns3/ew3 combination is omit-
ted. The direct, strong, and background fits determine the reported amplitude and
188
uncertainties. (F) The “best” Matlab method includes an average of the m4 pro-
cess with the ns3c/ew3z and ns3z/ew3z combinations. Again, the direct, strong, and
background fits determine the amplitudes and uncertainties.
The AX amplitudes are spread evenly within 10 aT of a zero value and the AY am-
plitudes have a tighter spread clustered around 30 aT. Each of the reported statistical
and systematic uncertainties are tightly grouped. Given the tight cluster of points,
an average of the results from methods A-F to a single amplitude pair is reasonable
and provides the following result
AX = (0.001± 0.019stat ± 0.010syst) fT (5.6)
AY = (0.032± 0.019stat ± 0.010syst) fT. (5.7)
The amplitude for AY is a 1.5σ result and is interpreted to be consistent with zero.
5.4.4 SME Parameters and Energy Scaling
The comagnetometer is sensitive to the difference in electron and nuclear spin cou-
plings (Eq. 3.9). The limit on an electron Lorentz-violating field from Ref. [86] is
equivalent to a magnetic field of 0.002 fT and is more tightly constrained than Eqs. 5.6
and 5.7. Therefore, we can ignore a possible signal from electron interactions and set
clean limits on Lorentz-violating nuclear spin interactions.
These results can be interpreted in terms of the parameters in the SME [9]. The
following 3- and 4-dimensional operators in the relativistic Lagrangian can be con-
strained from coupling to a spin-1/2 particle
L = −ψ(m+ bµγ5γµ +
1
2Hµνσ
µν)ψ +1
2iψ(γν + dµνγ
5γµ +1
2gλµνσ
λµ)←→∂νψ (5.8)
189
where bµ and gλµν are CPT-odd, while dµν and Hµν are CPT-even fields. Five- and
six-dimensional operators not included in the SME also lead to higher order spin
coupling terms [87, 88].The energy shift from the comagnetometer is
δE = −µ3He(−βni )σi = −µ3HeAiσi (5.9)
where the negative sign on βn accounts for the response of the comagnetometer to an
anomalous nuclear spin coupling and σ is the Pauli matrix8.
The 3He nucleus contains one neutron and two protons. To leading order, the
protons are paired in a singlet state leaving the nuclear magnetic moment to be
dominated by the neutron. The polarization of each nucleon has been calculated to
be Pn = +0.87 and Pp = −0.027 for the neutron and proton respectively [89]. The
leading energy energy shift of the 3He nucleus in the SME becomes
δE = −Pnbni σi − 2Ppbpiσi (5.10)
where b is a linear combination of SME coefficients9. In cartesian coordinates with
the quantization axis along the z-axis,
bn3 = bn3 −mndn30 +mng120 −Hn
12 (5.11)
where mn is the mass of the neutron [17].
Given these relationships, the following limits can be placed for the neutron
bnX = (0.1± 1.6)× 10−33 GeV (5.12)
bnY = (2.5± 1.6)× 10−33 GeV (5.13)
8Ai is defined as in Eq. 5.2.9The factor of 2 comes from the multiplicity of 2 protons.
190
r
x
y
y0 ±
σy
x0±σx
Figure 5.19: Two-dimensional gaussian interpretation of a limit given the uncertainty σxand σy. A radius r that encloses 68.3% of the enclosed area determines the limit at the 68%confidence level.
using the fact that µ3He = −6.707× 10−17 GeV/T. This result represents the highest
energy spin anisotropy measurement to date.
5.4.5 Interpretation to a Limit
If the final amplitudes reported in Section 5.4.3 were zero, the limit on |b⊥| = |b2X +
b2Y | would be simple to determine from propagating the uncertainty. The nonzero
amplitudes influence the smallness of the limit especially the 1.5σ amplitude for bY .
A simple gaussian interpretation of the parameters can set a conservative limit on
b⊥. For example, given the measurements bX = x0± σx and bY = y0± σy, the central
values are offset from the origin with 1σ uncertainties σx and σy (Fig. 5.19). A limit is
determined by integrating the probability within a fixed radius centered at the origin
that contains 68.3% of the two-dimensional gaussian,
0.683 =1
2πσxσy
∫ 2π
0
∫ R
0
e(−r cos(θ)−x0)2/(2σ2x)e(−r sin(θ)−y0)2/(2σ2
y)rdrdθ (5.14)
where R determines the radius of the circle that constrains the total integral. This
method produces a limit of |bn⊥| < 3.7× 10−33 GeV at the 68% confidence level10.
10In other units, this result is equivalent to 49 aT or 1.5 nHz.
191
b
10-33
10-32
10-31
10-30
10-29
10-28
Lim
its
on |˜
n ⊥| (
GeV
)
Bear et al. (2000)Kornack (2005)Altarev et al. (2009)Brown et al. (2010)Gemmel et al. (2010)
Har
var
d-S
mithso
nia
n129X
e/3 H
e m
aser
CP
T-I
K-3H
e co
mag
net
omet
er
ILL, G
renob
leU
ltra
cold
Neu
tron
s
CP
T-I
IR
ota
ting
K-3 H
e co
mag
net
omet
er
Main
z3 H
e/129X
e co
mag
net
om
eter
Figure 5.20: Summary of published measurements of bn⊥ in the last decade.
Others in the field, for instance Ref. [40], follow the same method as Eq. 5.14; how-
ever, they assume that both x0 = 0 and y0 = 0 in determining the radius of the limit.
This method represents only the precision of the measurement, not the measurement
itself. In comparing the work of this thesis with other results, all measurements have
been standardized to the two-dimensional gaussian approach.
5.4.6 Competitors
Several experiments have performed searches for bn⊥ over the last 10 years (Fig. 5.20).
The Harvard-Smithsonian noble-gas maser limit stood as the most precise measure-
ment of bn⊥ for a decade although the results from CPT-I achieved comparable sensi-
tivity. In 2009, work in Grenoble using ultracold neutrons presented a less-sensitive
limit. Despite the lower sensitivity, the ultracold neutron measurement is particularly
interesting because it is the first limit on the unbound neutron. The result is com-
pletely independent of nuclear structure and unaffected by model-dependent nuclear
corrections. It is also free from possible bound state suppression effects. In addition,
the measurement surpasses the sensitivity to the free proton by a factor of 200 [35]
to present the best free nucleon limit.
192
The Harvard-Smithsonian dual noble-gas spin-maser uses the precession of two
polarized noble-gas species to search for a sidereal variation in the ratio of 129Xe/3He
precession frequencies. In a static field of 1.5 gauss, the spins precess at 4.9 kHz and
1.7 kHz for 3He and 129Xe respectively. The quantization axis of the spins is oriented
along the E-W direction for optimized spatial overlap with the equatorial plane. An
inductive pickup coil measures the spin precession of both species and the precession
of the 129Xe is compared against a H-maser. A feedback loop stabilizes the magnetic
field leaving the 3He free to precess. A sidereal oscillation in the precession frequency
of 3He is a signature of a Lorentz-violating field.
The 129Xe/3He maser has similar short term sensitivity to the K-3He comagne-
tometer but is also able to extend this performance to lower frequencies than in the
best noise spectrum in this work (Fig. 3.15). This demonstrates incredible long term
stability and suppression of low frequency noise. Using the values listed in Table 1 of
Ref. [29], the sidereal amplitudes of each phase are determined11 to be
bnX =(+5.5± 5.7)× 10−32 GeV (5.15)
bnY =(−0.6± 5.4)× 10−32 GeV. (5.16)
This result is limited by laser and temperature stability over the course of a day.
Placing the 129Xe/3He maser on a rotating platform and performing frequent reversals
would not be as beneficial as it has been for the K-3He comagnetometer. Both the
129Xe and 3He have long coherence times of hundreds of seconds which limits the
frequency of reversals to this timescale.
A group in Mainz, Germany also use a comparison of the spin precession of 3He
and 129Xe in a co-located vapor cell as a 3He/129Xe comagnetometer [40]. Both
species are optically pumped in spherical vapor cells which are then placed as close as
11The similar computation in Ref. [28] fails to take into account the reversals from east and westorientations of the magnetic field.
193
possible to the SQUID in a magnetically shielded room. A tipping pulse excites the
spins into the transverse plane in a guiding magnetic field of 400 nT. The signal is a
superposition of both the ω3He = 2π × 13.4 Hz and ω129Xe = 2π × 4.9 Hz precession
measured with a low-Tc DC-SQUID magnetometer. From a fit to the free induction
decay, a phase difference
∆Φ(t) = ΦHe(t)− (γ3He/γ129Xe)ΦXe(t) (5.17)
removes magnetic field effects where the value of γHe/γXe is known to a part in 109. A
Lorentz-violating field leads to a sidereal oscillation in ∆Φ(t) while Earth’s rotation
rate leads to a linear phase drift. The Mainz 3He/129Xe comagnetometer places the
following bounds
bnX =(+3.36± 1.72)× 10−32 GeV (5.18)
bnY =(+1.43± 1.33)× 10−32 GeV. (5.19)
in 7 runs totaling less than 140 hours. Each run is limited by the 129Xe transverse
relaxation time to 8-16 hours by wall relaxation. In principle, longer times are possi-
ble. Due to the coherence of the spins, the statistical uncertainty of a run decreases
as 1/T 3/2 rather than 1/T 1/2 for independent measurements. This experiment will
not benefit from a rotating platform as the coherence times are tens of hours.
Using the method in Section 5.4.5, bounds on |bn⊥| can be placed for each experi-
ment consistent with the interpretation in this work. The Harvard-Smithsonian work
sets |bn⊥| < 1.1 × 10−31 GeV, and the Mainz results sets |bn⊥| < 4.7 × 10−32 GeV. If
the wall relaxation times of 129Xe can be extended in the 3He/129Xe comagnetometer,
these results will be the most direct competitor to this work.
194
5.4.7 Beyond bn⊥
The analysis presented here constrains bX and bY using the signature of a sidereal
oscillation. These results are combined to constrain |bn⊥|. Other measurements are
possible with the measurements of AX and AY , but have not been strongly pursued.
Future improvements to this experiment will continue to increase the sensitivity to
these and other parameters. Final limits on a wider variety of possible parameters
will be determined from an optimized experiment. Some of these parameters are
discussed below.
The parameter bZ is unconstrained for both the proton and the neutron [27]. A
method has been proposed to use axial precession of the Earth to take advantage of
a small sidereal oscillation of bZ [90]. The Earth’s axis has been observed to precess
about the normal to the ecliptic plane over a period of 25,771.5 years. The presence
of a nonzero bZ would appear as a slow linear drift in the AX parameter. It is clear
from the data that such an analysis will place a limit on bZ that is not as stringent
as the limits placed on bX and bY . Combining the Harvard-Smithsonian and the
Princeton data should improve this result from a 10-year baseline between each set
of measurements. This collaboration is in progress under the direction of the author
of [90].
The measurements of AX and AY can also constrain theories of spacetime torsion
[91]. Gravity curves spacetime with energy-momentum density as its source, but
torsion twists spacetime with some models predicting a spin density as its source.
The effects of torsion introduce a preferred orientation for a freely falling observer
as is the case for local Lorentz violation, so this limit can also be interpreted as a
limit on spacetime torsion [91]. Similarly, continued measurements extending beyond
a year would more tightly constrain the boost invariance of the neutron [37]. The
signature of this effect is an annual variation in the daily sidereal modulation of AX
and AY .
195
CPT-II CPT-IExisting Limit Source
(This Work) (2005)
|bn⊥| 3.7× 10−33 GeV 1.4× 10−31 GeV 1.1× 10−31 GeV 3He/129Xe-Maser[36]
|bp⊥| 6.0× 10−32 GeV 2.2× 10−30 GeV 1.8× 10−27 GeV H-Maser[35]
|be⊥| 2.8× 10−30 GeV 1.0× 10−28 GeV 1.5× 10−31 GeV Torsion Pendulum[86]
Table 5.3: Lorentz Violation limits for the neutron, proton, and electron as determined byCPT-I and CPT-II compared with the best limits on each particle.
The nonzero contribution of the proton to the 3He nucleus (Eq. 5.10) allows a
limit of |bp⊥| < 6.0 × 10−32 GeV. This is the most sensitive measurement for proton
Lorentz violation [27]. A limit on the electron Lorentz violation from this work is not
nearly as stringent as the best limit (Table 5.3). However, this result constrains the
electron contribution to the comagnetometer response to set clean limits on nuclear
spin couplings. The proton limit can also be used in combination with the 3He-129Xe
maser result [29] and recent analysis of the nuclear spin content of 129Xe [92] to set
an independent stringent limit on proton Lorentz violation.
There are many models involving higher order operators not considered within
the SME [87, 93, 94], and limits on these parameters are also improved. Even though
only spatial components of spin anisotropy in the Earth’s equatorial plane have been
presented here, time-like components of Lorentz violation could also be observed
since the solar system is moving relative to the rest frame of the cosmic microwave
background (CMB) with a velocity v ∼ 10−3c at a declination of −7. Possible
anisotropy associated with the alignment of low-order CMB multipoles also points in
approximately the same direction [95].
5.4.8 Proposed Improvements
The limit presented here could be further improved by continued measurements fol-
lowing the procedure developed by the end of this collection. An appropriate calibra-
tion using the low frequency Bx calibration throughout the entire dataset as well as
196
implementation of the probe background measurements would provide lower scatter
in measured sidereal amplitudes (Fig. 5.17). The 143 days of data collected over nine
months have improved the existing limit by a factor of 30. Continued measurements
with this scheme would improve the uncertainty by the square root of the number of
days, but would likely be limited by any finite amplitudes measured.
During the completion of this work, an improved alkali-metal noble-gas comagne-
tometer has been developed by replacing 3He with 21Ne [74]. The gyromagnetic ratio
of 21Ne is an order of magnitude smaller than 3He resulting in an order of magni-
tude greater energy sensitivity for the same magnetic field background. The spin-3/2
nucleus of 21Ne also contributes a significant quadrupole moment. This leads to in-
creased relaxation of the nuclear spins on the order of 90 min over the case of 3He, so a
hybrid alkali optical pumping scheme has been implemented [96] to achieve a compen-
sation frequency of 10 Hz. The pump laser pumps an optically thin K vapor without
significant pumping efficiency loss by absorption through the cell. The polarized K
more uniformly polarizes an optically thick Rb vapor for more efficient pumping of
the 21Ne. A probe laser near 795 nm samples the response of the Rb atoms. This K-
Rb-21Ne comagnetometer can provide at least an order of magnitude improvement in
sensitivity to nuclear spin couplings. The details of this comagnetometer are beyond
the scope of this work.
This K-Rb-21Ne comagnetometer has an order of magnitude greater sensitivity to
rotations as well as anomalous nuclear spin couplings. Rotational effects that can be
safely neglected in this thesis will play a role in the next generation of experiments.
To solve these issues, this work further proposes that the experiment be placed at
the South Pole for removal of the Earth-fixed gyroscope background. At either the
North or South Pole, the gravity vector and the rotation vector align for suppression
of both tilt and gyroscope effects. A landmass and research station are available at
the South Pole making this location more attractive.
197
Placement of the experiment 100 m from the South Pole will yield 54 aT as the
maximal Earth’s rotation amplitude even with the enhanced sensitivity of the K-Rb-
21Ne comagnetometer. The total integrated limit from this work is 49 aT, so Earth’s
rotation effects can almost be neglected at this location. Furthermore, all of the
systematic effects relating to the N-S and E-W reversals described earlier are removed.
The Ωz calibration method in Section 4.3.4 will still provide an important check of the
calibration. If systematic effects in a reversal can be sufficiently controlled, it would
also be possible to reverse the experiment along the celestial X- and Y-axes instead
of choosing a fixed Earth-frame direction and searching for a sidereal oscillation in
the response. This no longer limits new measurements to the period of one day and
ultimately leads to faster averaging of the signal.
Plans for an experiment at the South Pole are currently underway. Development
of the K-Rb-21Ne comagnetometer on the rotating apparatus has recently placed
a new limit on tensor Lorentz violation based on the methods of this thesis [97].
Future work will include upgrades to the CPT-II apparatus for more rugged and
automated operation including more sensitive tilt measurements and replacement of
the platform with a more rugged air bearing. Experiments and development will
continue in Princeton, NJ before shipment of the apparatus to the South Pole. The
final measurement is expected to achieve energy sensitivity on the order of 10−36
GeV (1 pHz) where effects suppressed by two powers of the Planck mass such as
dimension-6 Lorentz-violating operators [88] can be observed.
198
Chapter 6
Nuclear Spin Source
Hyper-polarized 3He targets are a well-developed technology for both atomic [98]
and nuclear experiments [99, 100]. A nuclear spin source containing 9 × 1021 hyper-
polarized 3He spins has been constructed for tests of long-range spin-dependant forces
using a K-3He comagnetometer. The design of this spin source incorporates accurate
polarization measurements and frequent reversals of the 3He spins. Optimization of
the reversals by adiabatic fast passage has achieved losses of less then 2.5 × 10−6
per flip. This is key to maintaining a high polarization under frequent reversals.
Details in this chapter will be specific to the design and construction of the spin
source1 to complement details regarding the execution of a new limit of long-range
spin-dependant forces using the CPT-I comagnetometer in Refs. [42, 43].
6.1 Spin Source Design
For a test of a long-range spin-dependant force, a spin source is placed outside the
magnetic shields of the CPT-I comagnetometer (Fig. 6.1). The spins in the spin
source are reversed every 2.8 s. A correlated response on the comagnetometer will
indicate the presence of a spin-spin interaction between the spins in the source and
1Some of these details have been described in an earlier report [101].
199
Oven
P
Pump Laser
P
P
SP
Probe Laser
PEMλ /4
P
LCW
Spin Source
Mag
net
ic s
hie
lds
LDA
Coils
λ/4
PD
z
x
y
λ/4
OA
PMF
PD
Intensityfeedback
Τ
Figure 6.1: CPT-I comagnetometer installation with spin source mounted vertically tosearch for an ~S1 · ~S2 interaction. PD: photodiode, SP: stress plate, T: translation stage, P:polarizer, PMF: polarization maintaining fiber, OA: optical amplifier, LCW: liquid crystalwaveplate, PEM: photoelastic modulator, λ/4: quarter waveplate. LDA: laser diode array.Image Credit: [42].
the spins in the comagnetometer. The sensitivity of the measurement scales with
the number of the spins in the source and increases with proximity of the spins in
the source to those in the comagnetometer. Several spin-dependant potentials are
proposed with either a ~S1 · ~S2, ~S1 × ~S2, or 3(~S1 · r)(~S2 · r)− ~S1 · ~S2 interactions with
several different inverse scalings with the distance [42]. An optimal spin source for
this experiment has a high density of polarized spins and a compact design. This
allows for a search for both ~S1 · ~S2 and ~S1 × ~S2 type interactions by reorienting the
axis of the spin source.
6.1.1 Early Spin Source Considerations
A hyper-polarized liquid 129Xe spin source would achieve a higher density of polarized
spins, but the total volume of liquid 129Xe is restricted. In addition, placement of
a cryogenic spin source as close as possible to the comagnetometer shields presents
obvious engineering challenges. A greater number of polarized spins can be developed
200
in a high volume, high density 3He vapor cell. High pressure vapor cells are subject to
unexpected and catastrophic failure, and a practical limit of a 20 atm gas at operating
temperature has previously been observed [102]. Several key features were modeled
prior to construction to determine the specifications given this constraint.
6.1.2 Description
Figure 6.2 provides a schematic of the hyper-polarized 3He spin source. A high power
laser diode array bar containing 20 individual diodes provides several watts of pump-
ing light. An external cavity with internal lenses in a Littrow configuration adjusts
the wavelength to 770.1 nm and narrows the spectrum to a width of 0.1 nm where the
pressure broadened linewidth in K is 0.3 nm. The spectrum is continuously monitored
using an integrating sphere and spectrometer for occasional cavity adjustments. A
liquid crystal waveplate allows for automated control of the handedness of the pump-
ing light. Several cylindrical lenses and mirrors reshape the beam and deliver 2 W of
pump light to the cell.
The cylindrical vapor cell is filled with 12.0 amagats of 3He, several mg of K,
20.4 Torr of N2 and measures 12.8 cm long with an inner diameter of 4.3 cm. The cell
is heated to 190 in a cylindrical G-7 oven by forced hot air. Pyrex windows sealed
with Viton o-rings provide optical access along the axis, and the entire assembly
is surrounded by an insulating polyamide foam. The oven also supports three axis
magnetic field coils (Section 6.2.3). Several windings on top of the insulation form an
anti-Helmholtz coil for control of the gradient along the axis. A mirror retroreflects
pump light back onto the cell and a photodiode monitors the transmitted intensity.
A fluxgate next to the oven monitors the magnetic field of the spin source. The entire
assembly is mounted on an 18 × 24 inch optical breadboard for mounting near the
201
LD
A
Grating
ISMultimodeFiber
Spec
trum
Anal
yze
r
FG
Mirrors
Mirror
λ/
LCW
Lenses
PD
Pol
Oven
Coils
Insu
lation
BSCC
Cel
l
Enclosure
BS
EPR
Figure 6.2: Schematic layout of the spin source. LDA: Laser Diode Array, Pol: Polarizer,LCW: Liquid Crystal Waveplate, FG: Fluxgate, CC: Compensation coil, EPR: EPR Coil,BS: Beam Sampler, PD: Photodiode, IS: Integrating Sphere. Image Credit: [43].
comagnetometer. A sheet metal and cardboard enclosure painted flat black prevents
photons from the spin source from reaching the comagnetometer2.
6.1.3 Design Considerations
The dimensions of the cell are determined by a model predicting the number of
polarized spins under a variety of conditions. The model considers spin-exchange
optical pumping by either K or Rb and predicts either 3 W or 60 W of laser power
respectively to polarize 1022 3He nuclear spins in a 200 cm3 cell at a maximal pressure
of 20 atm at operating temperature. A supply of several laser diode array bars each
capable of supplying several watts of power at 770 nm were available in the lab, so
K was selected. In hindsight, it would have been useful to consider hybrid optical
pumping, as these cells achieve higher polarizations [103].
2Early in the experiment, photons from the spin source reached an intensity feedback for thepump beam on the comagnetometer to produce a correlated response on the comagnetometer withreversals of the spin source.
202
The model determines the total number of polarized spins as a function of tem-
perature, 3He density, laser power, and aspect ratio of a cylindrical cell. The model
considers optical pumping on resonance of a pressure broadened absorption line in
Eq. 2.5 and follows from the propagation model in Section 2.1.4. Spin-exchange
cross sections are extracted from measured spin-exchange rate constants provided in
Refs. [104, 105] to determine the rates at different temperatures (Section 2.2). The
3He dipole-dipole relaxation rate in Eq. 2.23 is modified to include a temperature
dependance as
1
T dd1
=pn
744 amagat · hour
√296 K
T(6.1)
which holds to a good approximation below 550 K [55, 106].
The model also considers an observed limit to spin-exchange optical pumping cells,
the X-factor, as a modification to Eq. 2.25
P bz = 〈P a
z 〉Rnese
Rnese (1 +X) +Rn
sd
(6.2)
where X depends on the surface area to volume ratio of the cell [54]. The plots in
Fig. 2 of Ref. [54] provide an estimated value X = 0.3 for S/V = 0.45 of the cell
used in the experiment (Section 6.1.4). Given all of these factors, the model predicts
a 40% polarization (2× 1022 spins) for 3 W of pumping light in a 200 cm3 vapor cell
at 190C. In practice, only 20% polarization is achieved.
Alignment of the pump beam can limit the polarization in optically thick cells
through a process known as skew light optical pumping. By pumping atoms along
a direction misaligned with the quantization axis, the vapor is no longer transparent
and absorption diminishes the propagation of light through the cell. The absorption
by the alkali-metal vapor is shown to roughly double at an angle
θ ∼ 1√nalσ
(6.3)
203
over a distance l. For our cell, na = 9.5 × 1013/cm3, l = 12 cm, and σ = 3.5 ×
10−12 cm2 predicts an angle of 9. Given the geometry of the optical layout, the
beam alignment is restricted far below this level and it is unlikely that this effect is to
blame. In addition, it is strongly suspected that pumping of the D2 line also limits the
polarization [107]. Experienced individuals in this field recommend using much more
laser power than the model indicates to achieve high polarizations [108]. Despite the
factor of 2 difference between predicted and measured polarizations, 9×1021 spins in a
compact source that can efficiently reversed has been achieved for a test of long-range
spin-dependant forces between nuclear spins.
6.1.4 Cell Manufacture
Three cells were hand blown from a 52 mm stock tube of Corning 1720. This glass
has not been manufactured for 20-30 years and has been gathering defects while in
storage. These defects can lead to catastrophic failure in a high pressure cell, so this
tube is resized to freshly melt the glass and remove defects. Remelting the glass also
increases the wall thickness to 4 mm for added strength. Failure of a cylinder is most
likely along the axis from the hoop stress3 given by
σh =pφ
2t(6.4)
in the limit where φ t where p is the pressure inside the cylinder, φ is the diameter,
and t is the wall thickness. Sharp corners at the faces of the cylinder weaken the glass
so the ends are curved outward4. There is little distortion of an image as viewed along
the axis of the cylinder.
3This is why a boiled hot dog splits along the axis, never along the radius.4Curving the windows in like the bottom of a wine bottle will lead to a more concentrated force at
the joint and reduced structural integrity. It also further distorts the optical quality of the windows.
204
The 4 mm wall thickness is selected using the practical and design limits of glass at
a given pressure. Glass suppliers recommend a design limit of 0.1 inch wall thickness
for a cylindrical cell with inner diameter of 1 inch containing 200 psi [109]. Using
Eq. 6.4, this recommends a wall thickness of 7 mm for a 12 amagat (300 psi at 190C)
cell. The design limit is set an order of magnitude below the known breaking point of
the glass, but the practical limit is only a factor of 2 smaller than the breaking point.
The cell is maintained under ideal conditions since it is evacuated and filled with
a dry inert gas where all the water has been driven off when heated. Operation of
cell between the practical and design limits provides an acceptable trade off between
volume and structural integrity.
Any remaining defect can cause the cell to explode, and an explosion from a cell
containing 20 atm of helium at 200C is equivalent to 250 mg of TNT. Each of the
three cells are hydrostatically pressure tested prior to filling with 3He gas to avoid
this issue. The cells are filled with and immersed in distilled water. Compressed
nitrogen pressurizes the cell. The criteria that a cell is acceptable for filling is that it
maintains a pressure of at least 25 atm for at least 45 minutes. This is 5 atm beyond
its intended operating condition. Two cells passed, so a third was tested beyond the
specification. It exploded after 45 minutes at 27 atm followed by 5 minutes at 28 atm.
These cells are filled with 12.0 amagat of 3He using the procedure in Ref. [43], but
one failed at the pulloff during initial heating.
The surviving cell has an inner diameter of 43 mm, outer diameter 52 mm, in-
terior volume 186 cm3, and length 13.6/12.5 cm with/without the curved windows5
(Fig. 6.3a). This cell provides the results in all measurements in this chapter. Due
to the high pressure even at room temperature, it is best considered a bomb. Great
care is always used when handling the cell. Fragments from the broken cell are large
5An average length of 12.8 cm is assigned assuming the cell is a regular cylinder.
205
a)
b)
c)
15 cm
15 cm
Figure 6.3: a) High pressure 3He cell. A magnetic coil is wrapped around the cell to mimicthe magnetization of the polarized 3He gas. b) Exploded cell fragments. c) The highpressure cell is treated as a bomb.
and extremely sharp (Fig. 6.3b), so protective clothing is always worn when handling
the cell (Fig. 6.3c).
6.2 Efficient Spin Reversals
Adiabatic fast passage (AFP) provides a means of efficiency reversing spin orientation
by sweeping a transverse, oscillating magnetic field through the resonance determined
by ω0 = γnBz. This technique is more effective than a NMR π/2 pulse if there are
inhomogeneities in the applied field [61]. A liquid crystal waveplate reverses the
handedness of the circularly polarized pump light for continuous pumping in either
orientation.
6.2.1 Adiabatic Fast Passage
For spins along the z-direction, an oscillating magnetic field along the transverse
direction By = Brf cos(ωt)y produces two counter-rotating fields at ω and −ω. In a
206
frame co-rotating with one of the components,
~Beff = (Bz −ω
γn)z +
Brf
2y (6.5)
where the effect of the counter-rotating field has been ignored. When the frequency
of the oscillating field equals ω0 = γnBz, the effective field along z vanishes and the
spins precess about y. An oscillating field at a frequency that sweeps through the
resonance at a rate slower than the spin precession will reverse the spin orientation.
For efficient AFP, Ref. [110] indicates that the following inequality should be well
satisfied,
D3He
|~∇Bz|2|Brf |2
ω
γnBrf
γnBrf (6.6)
where the term to the left is the 3He gradient relaxation in Eq. 2.24 expressed in
terms of the sweep at resonance. Typically Brf < Bz, so the relaxation rate is largest
at the resonance. This relaxation rate sets the criteria for the fast condition while
the term on the right describes the adiabatic condition. To decrease the lower bound,
substantial effort was placed in improving the homogeneity of the magnetic fields over
the entire region of the cell (Section 6.2.3).
6.2.2 Simulation
To determine efficient AFP parameters, a numerical simulation compares different fre-
quency sweeps. The 3He polarization is modeled in the absence of gradient relaxation
using
d~P n
dt= γn ~B × ~P n (6.7)
where Bz = B0 and By = A(t)Brf cos(φ(t)). A(t) scales the amplitude of Brf and φ(t)
describes the frequency sweep where several variations are considered. Efficient flips
are produces from a symmetric trapezoidal amplitude function and a linear frequency
207
sweep. For a sweep over time τ , these parameters are described by
A(t) =
t/τ1 : 0 ≤ t ≤ τ1
1 : τ1 < t < τ − τ1
τ−tτ1
: τ − τ1 ≤ t ≤ τ
φ(t) = ωst+1
2at2 (6.8)
where τ1 is the time over which the amplitude is ramped6, ωs is the starting frequency,
and a is the sweep rate. The factor of 1/2 in φ(t) is important to accurately describe
the linear frequency sweep.
In simulations, the most efficient reversals occur for Brf = B0/2, τ1 = τ/4, and
a linear frequency sweep between ω0/2 and 3ω0/2 over 1 s. This achieves losses of
4 × 10−6 per flip. Sweeping over a broader range quickly increases the losses to the
1% level. Adjustments to τ1 are more robust as long as τ1 6 τ . Similar results hold
for Brf ' B0. For reversals every 10 s, we expect an additional relaxation mechanism
1
TAFP1
' 1
700 hours(6.9)
that can be added to Eq. 2.22. The cell is limited by the dipolar 3He collisions
with T1 ' 70 hours, so these reversals will result only in a modest reduction in
the polarization of the cell. It is expected that the implementation of such a sweep
will result in lower efficiency because the model does not account for the effects of
magnetic field gradient relaxation.
6.2.3 Magnetic Field Optimization
It is not clear based on the limit in Eq. 6.6 how uniform the magnetic field needs to be
in order to achieve efficiency similar to the model. Ultimately, we attempt to create
6For numerical simulation purposes, A(t) is expressed in terms of hyperbolic tangent functionsto provide continuous derivatives.
208
Coil Calibration UniformityBx 0.688 gauss/A -By 3.61 gauss/A < 10−2
Bz 4.10 gauss/A < 5× 10−3
Table 6.1: Measured coil specifications closely match the model.
a static Bz field and transverse Brf that are as uniform as is reasonable to achieve.
The field of a finite solenoid does not meet the requirements of this application as it
would be too long for a compact spin source.
There are several magnetic field optimization schemes in the literature. An at-
tractive approach appears to be an evaluation of the magnetic field over an arbitrary
region from many current carrying conductors that can be cast into a linear program-
ming problem [111]. This approach is useful in applications where thousands of turns
are used but is impractical for our compact spin source. A more appropriate approach
is proposed in Refs. [112, 113] to use increasing numbers of current loops carrying
the same current that are appropriately placed to cancel higher order derivatives of
the magnetic field at the origin. This approach can be extended to reduce the total
gradient across the spin source7.
This sophisticated method was vetoed, so numerical evaluation of 28 discrete rings
based on uniform spacing were adjusted to reduce the overall gradient. This forms
the Bz field. Uniform transverse fields based on a cylindrical geometry are produced
by cosine windings defined in Appendix D of Ref. [114]. The angle of the jth wire in
a quadrant is given by
φj = arcsin[ 1
N
(j − 1
2
)](6.10)
where N = 7 for the By coil and N = 4 for the Bx coil. The specifications of each
coil are listed in Table 6.1.
7This technique has been implemented by a local company Twinleaf.
209
The coils are hand wound on grooves in a 11.5 inch long 3.190 inch outer diameter
G-7 tube (Fig. 6.4). The walls are 0.25 inch thick to allow windings of all three coils
at different depths from 22 gauge wire. The Bz coil contains trigonal windings: two
wires laid side by side with the third centered on top for each ring. Wires connecting
the individual rings form a twisted pair to reduce their effect on the homogeneity.
The form has two 0.5 inch holes for air intake and outake of the forced hot air heater.
Loops avoiding these holes are considered in the model. The coil form is insulated
with polyamide foam and polyamide (Kapton) tape to reach an outer diameter of
5 inches and is supported by two G-10 posts.
6.2.4 Performance
The AFP sweep is produced by a National Instruments PCI-6229 multifunction card
and digitally controlled by a computer. The signal is digitized at 833 kHz, and
a simple low pass filter at 100 kHz smooths discontinuities. In a holding field of
Bz = 7.8 gauss, a sweep from 12-36 kHz with τ = 80 ms, τ1 = 30 ms, and Brf =
0.5 gauss produces a repeatable reversal with losses less than 2.5×10−6 per flip. This
is evaluated by measuring the losses over 3000 consecutive reversals. The amplitude
Brf is limited by the Pasco Scientific 9587C amplifier and the inductance of the By
coil. The high efficiency of the AFP is attributed to the uniformity of the magnetic
fields. Despite close placement to the magnetic shields of the comagnetometer where
magnetic flux is concentrated, the reversals still maintain a high efficiency without
significant improvement from external gradient compensation. This result surpasses
a loss rate of 2× 10−5 per flip reported in Ref. [115] by an order of magnitude.
210
a)
b)
c)
15 cm
15 cm
Figure 6.4: a) Magnetic field coils neatly wound on the G-7 coil form. b) Oven withsupporting posts. A groove at each end supports circular pyrex windows (not AR coated)and seals with a Viton o-ring. c) Three dimensional rendering of the coil form design inSolidWorks.
211
6.3 3He Polarimetry
The strength of the anomalous interaction is proportional to the number of polarized
spins in the source. An accurate measurement of the 3He polarization can be imple-
mented using a polarization-dependant frequency shift in the alkali Zeeman resonance
as described in Ref. [110]. Effectively, the K atoms are used as a magnetometer to
measure the magnetization of 3He in real-time.
6.3.1 Electron Parametric Resonance
The 4S1/2 ground state splits into two hyperfine manifolds, F = 1 and F = 2. The
energy of each of the mF levels in the F = I± 12
state in a magnetic field of magnitude
B is described by the Breit-Rabi equation [44]
E(B,mF ) = − hνHFS
2(2I + 1)− gIµnBmF ±
hνHFS
2
√1 +
4mFxB2I + 1
+ x2B (6.11)
where xB = gJµBB/(hνHFS), νHFS = 461.7 MHz, gI = 0.3914/I [46], and µn is the
nuclear magneton. These levels are pictured in Fig 2.6b.
Circularly polarized photons pump the atoms to the end states with maximum
projection of angular momentum, |F = 2,mF = ±2〉. An RF magnetic field depopu-
lates the end state by driving atoms to the nearby |F = 2,mF = ±1〉 Zeeman sublevel.
These atoms are able to absorb photons which results in a dip in the transmission of
the pump light through the cell (Fig. 6.5). Exciting this transition is called electron
parametric resonance (EPR).
6.3.2 Polarization Measurement
An AFP flip reverses the magnetization of 3He and the local magnetic field experi-
enced by the K atoms. This corresponds to a shift in the EPR resonance of twice the
212
5.1 5.15 5.2 5.25 5.3
0.35
0.40
0.45
0.50
0.55
0.60
RF Frequency (Hz)
Tra
nsm
issi
on (
arb. units)
0 2 4 6 8 10 120
100
200
300
400
500
600
700
800
B (gauss)
EP
R F
requen
cy D
iffe
rence
(kH
z)
a) b)
Figure 6.5: a) Transmission dip from a frequency sweep across the EPR resonance. b)Difference in EPR resonance for opposite circular polarizations of the pump light as afunction of magnetic field.
magnetization. The frequency shift is related to the polarization P as
∆ν =8π
3
dν(F,mF ,±)
dBκµ3HenbP (6.12)
where ν(F,mF ,±) is the frequency of the EPR transition |F,mF 〉 → |F,mF ± 1〉
[110]. Due to the cylindrical geometry of the cell, the enhancement factor is modified
to
κ = κ0 +1
2− 3
2
(1 +
a
l−√
1 +(al
)2)
(6.13)
where a is the radius and l is the length of the cylindrical shell as reported in Ref. [104].
κ0 is as listed in Eq. 2.54 and produces a value of κ = 6.09 for our cell geometry at
190C.
The nonlinear part of the Breit-Rabi splitting implies that the EPR frequency will
be different depending on the handedness of the pump light. Figure 6.5b illustrates the
measured frequency difference as a function of magnetic field. The curve is determined
by Eq. 6.11 with no free parameters. Typical shifts from the 3He are 50 kHz, so this
is a significant effect. In a field of 7.68 gauss with a polarization of 100%, shifts of
213
0 10 20 30 40 50 60 704.7
4.8
4.9
5.0
5.1
5.2
EP
R F
requen
cy (
MH
z)
Time (s)
Figure 6.6: Three complete reversals of the 3He without reversing the handedness of thepump light easily determines the polarization value. This is followed by several reversals ofboth the 3He and the pump handedness. The kink in the leading edge is from an imperfectshift by the analog boost. This illustration is completed with a repetition of the startingreversals.
127.9 kHz or 119.4 kHz are expected for spins in the mF = −2 or mF = +2 state.
In the case of Ref. [110], the nonlinearity is only a small systematic effect where the
hyperfine splitting in Rb is νHFSRb = 1012 MHz.
A single rectangular loop of wire 5 cm wide extends along the length of the cell.
This is driven by a voltage controlled oscillator on a Wavetek 166 function generator.
A lock-in amplifier applies a 0.7 kHz modulation at 250 Hz to the EPR frequency. The
output of the lock-in creates a derivative signal for feedback from a simple analog PID
loop. A relay shorts the integral feedback during a reversal of the spins coordinated
with a reversal of the handedness of the pump light. The opposite Zeeman transition
is separated by several hundred kHz, so an analog signal provides a boost of the dc
feedback output to reduce the distance to the peak (Fig. 6.6). The loop quickly re-
engages to follow the EPR frequency. The frequency of the EPR is then monitored
by a counter on the PCI-6229 card.
214
6.4 Magnetic Field Compensation
The magnetic dipole interaction provides a known coupling between the spins in the
spin source and the comagnetometer. A 20% polarization produces a local field of
6 mgauss. An estimated shielding of 103 from the comagnetometer and 105 from the
shields reduces the strength of the interaction. Given the 48.7 cm distance between the
spin source and the comagnetometer cell, the interaction is estimated at 0.05 aT. The
spin source is designed to be surrounded by a µ-metal magnetic shield to suppress
the magnetic field; however, a solenoid is placed around the cell to provide active
cancelation of the magnetic field during reversals (Fig. 6.3a).
6.4.1 Compensation Coil
A continuous solenoid of 28 gauge wire extends the length of the cell. The magnetic
field of the coil mimics the surface current of a uniformly magnetized object [116].
The solenoid closely approximates the field despite the 4 mm thick cell walls and the
curved face of the cell. Shim coils were considered to compensate for these differences
but were not implemented.
6.4.2 EPR Consideration
During the experiment, compensation of the 3He magnetization contributes to the
field experienced by the K vapor. This compensation field does not cancel the EPR
frequency shifts since the enhancement factor κ 6= 1. This is demonstrated in Fig. 6.7.
For effective polarization measurements, the handedness of the pump light as well as
the current in the compensation coil are maintained constant under AFP reversals.
The known field applied by the compensation coil is accounted for in Eq. 6.11 in the
EPR measurement.
215
0 20 40 60 80 100 120 140 1605.13
5.14
5.15
5.16
5.17
5.18
5.19
Time (s)
EP
R F
requen
cy (
MH
z)
Figure 6.7: Measured EPR frequency over several AFP flips without a reversal of the pumppolarization. After 40 s, the magnetic field compensation is engaged and later disengagedafter 110 s. Due to the enhancement factor κ 6= 1, there is a nonzero EPR shift during areversal of both the 3He and the coil current. The spikes are an artifact of shorting thecapacitor for the integral feedback of the PID.
6.4.3 Performance
A measurement of the polarization described in Section 6.3.1 determines the appro-
priate current to apply to the coil with a measured calibration of 30.1 mgauss/mA.
This reduces the correlated magnetic field by a factor of 10 as monitored by a fluxgate
placed near the cell. The coil is also useful in determining the magnetic field leakage.
Reversals of a large current through the coil mimics the uncompensated field of the
spin source. A correlation with a magnetic field and the comagnetometer limits the
magnetic field leakage to 4× 10−3 aT.
6.5 Spin Source Operation
Operation of the spin source follows the same scheme as reversals of the CPT-II ap-
paratus described in Chapter 5. Historically, this implementation inspired the control
sequence in CPT-II. The spin source and the comagnetometer are electrically isolated
and powered by separate lab ac phases. A separate computer handles all of the data
216
acquisition and control of the spin source. Most of the electronics and computer
control of the spin source reside in the next room with several BNC and an ac line
connections passing through a hole in the wall. The two computers are synchronized
using Dimension 4 software and a navy timeserver connection. A trigger from the
comagnetometer to begin flipping sequences is disconnected electrically through an
optocoupled signal.
6.5.1 Flipping Sequence
The spins and compensation coil current are reversed every 2.8 s with synchronous
control of the liquid crystal waveplate and feedback boost. The fluxgate and EPR
frequency are continuously monitored and recorded to a file. The spin source also
maintains the state of the spins by switching binary values saved in a file indicating
the spin state. This is particularly useful if there is a power failure as the last value
will always be saved.
A typical reversal proceeds as follows: The AFP parameters are loaded and the
waveform is calculated. A digital channel controls the relay to short the capacitor on
the integral feedback of the PID. The AFP waveform is generated while analog levels
of the feedback boost and compensating coil are smoothly transitioned at resonance.
A digital channel switches the state of the liquid crystal waveplate at resonance and
is also read on an analog input as a marker of the state in the files. At the end of the
AFP waveform, the capacitor is unshorted for the PID to reacquire the EPR peak.
In addition, a digital channel marks the time during which the AFP is executed and
recorded on an analog channel as a marker of the flip.
After a 200 s record with over 60 reversals, the comagnetometer pauses to perform
zeroing routines. During this time, the 3He spins are reversed without reversals of
other parameters to perform a polarization measurement. Based on the measured
polarization, the current for the compensation coil is determined. Occasionally, the
217
feedback will fail to lock to the EPR peak potentially disrupting an entire record with
the wrong current to the compensation coil. To account for this, the previous ten
polarization values are saved and compared against the most recent value. If there
is a discrepancy by more than 2.5 standard deviations of the recent value and the
mean of the previous values, the recent value is replaced by the mean. This has led
to robust operation with few failures.
6.5.2 Long-Range Spin-Dependant Forces Search
A brief summary of the results of the long-range spin-dependant forces search will
be summarized here. Familiar techniques using the 3-point overlapping string analy-
sis provide the reversal correlated amplitude of the comagnetometer signal with the
orientation of the spin source. Specific details regarding the CPT-I comagnetometer
conditions and the analysis are already well-explained in Refs. [42, 43] and will not
be repeated.
Data collected over a month sets a limit on an anomalous spin-dependant force of
0.05± 0.56 aT with a reduced χ2 of 0.87. This improves upon the previous limit by
a factor of 500 [98]. In addition, it is equivalent to measuring a frequency shift in the
3He precession of less than 18 pHz and represents the highest energy resolution of any
experiment. Systematic effects are checked by collecting several runs with the spins in
the spin source and the comagnetometer reversed with respect to the electronic levels
of each system. These collections give consistent results (Fig. 6.8). The measurements
within each run form a gaussian distribution. In addition, correlated amplitudes are
also binned by sidereal day to set limits on Lorentz-violating spin potentials. The
structure of 3He in Eq. 5.10 gives the sensitivity of the measurement to both the
neutron and the proton.
Several different spin-dependant potentials correspond to the existence of newly
proposed particles. For example, the coupling gp of a pseudoscalar boson φ with mass
218
0 5 10 15 20 25 30 35 40
-10
-5
0
5
10
15
Sidereal day
Sig
nalβ
y (a
T) 0
200
400
600
800
0 100 200-100-200 aT
n
0.0 0.2 0.4 0.6 0.8 1.0-15
15
βn y(a
T)
-15
Sidereal time (days)
0
Cou
nts
Figure 6.8: Spin-correlated measurement of βny for the spin source aligned in the y-direction.Each point represents the average correlation over approximately one day. Up and downtriangles indicate the orientation of the spin source while the filled and empty trianglesindicate the orientation of the comagnetometer. Top Inset: Within each run, the correlationfrom 200 s records follows a gaussian distribution. Bottom Inset: Correlations plotted andaveraged modulo one sidereal day shows no significant variation. Image Credit [42].
m to a fermion ψ with mass Mn can occur from either a Yukawa or derivative form
LY uk = −igpψγ5ψφ or LDer =gp
2Mn
ψγµγ5ψ∂µφ (6.14)
where both terms lead to the same single boson exchange potential [117]
V (~r, ~S1, ~S2) =g2p
16πM2n
[~S1 · ~S2
(mr2
+1
r3
)−(~S1 · r)(~S2 · r)
(m2
r+
3m
r2+
3
r3
)]e−mr (6.15)
where r represents the distance between the spins with ~ = c = 1. This particular
potential was originally derived in the context of one-pion exchange where the mass
of the pion is much less than the mass of other more strongly interacting particles
[118] and reduces to the familiar dipole-dipole potential for m = 0. In addition, the
dynamics of a Goldstone boson π associated with spontaneous breaking of Lorentz
219
symmetry have been suggested [119]. This boson would couple to fermions as
L =1
Fψγµγ5ψ∂µπ +
M2π
Fψγ0γ5ψ (6.16)
to leading order and have an unusual quadratic dispersion relationship ω = k2/Mπ.
The first term leads to a 1/r potential and the second term leads to a ~S ·~v interaction
where ~v is the velocity with respect to a preferred frame. The velocity-dependant in-
teraction is constrained by considering a sidereal variation in the spin-spin interaction.
Further discussion of these potentials and others are left to Refs. [42, 43].
6.5.3 Future Improvements
The spin-dependant forces measurement can be improved through the increased en-
ergy sensitivity of a K-Rb-21Ne comagnetometer [97]. Likewise, the inverse distance
scaling of the potentials means that the smaller shields of the CPT-II comagnetometer
lead to an additional improvement in sensitivity. The existing spin source would need
to be modified to operate in a vacuum. A likely first step would be a replacement of
the forced hot air heater with a twisted wire resistive heating scheme. The external
cavity laser would likely need to be outside the bell jar to allow for occasional adjust-
ments to the cavity. In addition, it is believed that the 3He polarization is limited by
the laser power, so additional lasers will likely lead to continued improvements in the
spin source.
220
Chapter 7
Conclusion
This thesis sets new limits on a Lorentz- and CPT-violating field coupling to the spin
of the neutron. The equatorial components of the proposed Lorentz-violating field
are measured to be bnX = (0.1± 1.6)× 10−33 GeV and bnY = (2.5± 1.6)× 10−33 GeV.
This can be interpreted as a limit on |bn⊥| < 3.7× 10−33 GeV, improving the previous
bound by a factor of 30. This represents the highest energy resolution of any spin
anisotropy measurement.
This measurement was accomplished through the use of a K-3He comagnetometer
on a rotating platform for lock-in modulation of the Lorentz- and CPT-violating field
on a timescale much shorter than a sidereal day. The rotating platform introduces
several background effects that have been investigated in this work. New and more
robust calibration schemes have been presented along with observation of novel fea-
tures of the K-3He comagnetometer. Several improvements to the experiment are
proposed, most notably placement at the South Pole for a suppression of Earth’s
rotation as a fixed background signal.
Furthermore, a hyper-polarized 3He nuclear spin source was designed, constructed,
and used in combination with the K-3He comagnetometer to perform a search for long-
range spin-spin interaction from newly proposed particles. The spin source achieves
221
9 × 1021 polarized nuclear spins that can be reversed frequently with losses below
2.5 × 10−6 per flip. This search improves previous bounds by a factor of 500 and
achieves an energy sensitivity of 18 pHz. This is the highest energy resolution of any
experiment.
Each of these experiments will benefit from the improved energy sensitivity of the
newly developed K-Rb-21Ne comagnetometer [97]. The alkali-metal noble-gas comag-
netometer has achieved energy sensitivity of 10−34 GeV [42], and future improvements
are expected to reach the level of 10−36 GeV where effects suppressed by two orders
of the Planck mass can be explored.
222
Appendix A
Time Conventions
Lorentz violation searches present measurements in sidereal time [17]. The length of
a sidereal day is defined as the time it takes a point on the earth to rotate 360 with
respect to the celestial frame. A sidereal day contains 86164.09 s and differs from the
86400 s of a diurnal day from the combination of orbital and rotational motion of the
Earth with respect to an inertial frame. As a convention, all results are reported in
local sidereal time (LST).
The algorithm to calculate the sidereal time has already been implemented in
the first generation experiment and described in Ref. [28]. The calculation converts
the computer Unix time to solar days from January 1, 2000. These solar days are
then converted to sidereal days since January 1, 2000, to match the convention.
These times give the Greenwich Mean Sidereal Time (GMST) and are the values of
“Sidereal Day” reported in this thesis. The GMST values have been verified with
a known astronomy timeserver, and the rest of the calculation is confirmed to be
correct1.
GMST refers to the position of Greenwich, England, while the experiment occurs
in Princeton, NJ. The data is analyzed in LST to account for this difference. LST
1Automatic updates to the local time on the computer from daylight savings time will cause aphase shift in the recorded times, so this feature is disabled in all computers involved in experiments.
223
Z Z
YX X
χ
χ
nz
ny
nx
Right ascension
Celestial North Pole
Celestial Equator
Celestial FrameLab Frame
Vernal Equinox
Ecliptic
Earth
Declinatio
n
Ωt
Ω
Ωt
a) b)
Figure A.1: a) Transformation of Celestial Frame to Lab Frame as given in [17]. b) CelestialCoordinate System. Image Credit: [28].
and GMST differ by a correction in the longitude as
LST = GMST + λ/360 (A.1)
where λ is longitude of the observer on Earth. Princeton, NJ is at 74.652 W longi-
tude, so this correction is −0.207.
Measurements are reported in the sun-centered celestial equatorial frame (XYZ)
that is approximately inertial over thousands of years [17]. Axes are defined by
convention that is well established in the field of astronomy [120]. This coordinate
system assigns X to have a declination and right ascension of 0 in the equatorial
plane of the Earth (Fig. A.1b). Y has a declination of 0 and right ascension of 90.
Z is oriented along the Earth’s rotation axis, and these directions form an orthogonal
basis. For an observer at 0 longitude at the equator, the X direction is directly
overhead on an integer sidereal day.
224
In terms of the SME, the authors of Ref. [17] define the following coordinate
transformation2
nx
ny
nz
=
cosχ cos Ωt cosχ sin Ωt − sinχ
sin Ωt cos Ωt 0
sinχ cos Ωt sinχ sin Ωt cosχ
X
Y
Z
(A.2)
as seen in Fig. A.1a. Within this convention, the quantization axis is along nz.
The rotating comagnetometer is sensitive to fields interacting along its y-axis in the
horizontal plane. This axis can also be reoriented by rotation of the apparatus. We
define the following transformation to further rotate our coordinate system.
z
x
y
=
− sin θ cos θ 0
0 0 1
cos θ sin θ 0
nx
ny
nz
(A.3)
where θ is the angle of rotation lock-in amplifier from the south position as defined
in Appendix B. Under such a rotation, sensitivity to the Lorentz- and CPT-violating
field b becomes
y = bX(cos θ cosχ cos Ωt−sin θ sin Ωt)+bY (cos θ cosχ sin Ωt+sin θ cos Ωt)−bZ cos θ sinχ.
(A.4)
Rotations of θ determine the sensitivity of the reversal correlated amplitude for re-
versals along the N-S and E-W axes.
2Note that χ is used as the latitude elsewhere in the thesis where here χ − π/2 determines thelatitude.
225
Appendix B
Local Coordinate System
The CPT-II comagnetometer uses a rotating platform for measurement of possible
Lorentz-violating fields on timescales shorter than one sidereal day. The sensitive y-
axis of the magnetometer is oriented horizontally such that rotations of the platform
about the vertical z-axis reorient this direction with respect to the celestial frame.
The absolute orientation of the apparatus is a factor in determining the projection
of the measurement within the SME, as it was in the CPT-I experiment. However,
reorientation of the apparatus with respect to Earth’s rotation axis provides a signif-
icant Earth-fixed background response. Measurements of spatial anisotropy rely on
relative changes in this signal; however, a correct measurement of the gyroscope sig-
nal provides an important check of systematic effects on our signal. Here, we discuss
the determination of the location of CPT-II with respect to Earth’s rotation axis and
possible variations.
B.1 Laboratory Location
The comagnetometer is located at 40 20′ 43” N (40.345 N) and 74 39′ 7” W
(74.652 W) in the basement of Jadwin Hall at Princeton University in Princeton,
NJ. Since the y-direction of the CPT-II comagnetometer is mounted horizontally, the
226
Ω
40°20'43"N
y
z
ˆ
ˆ
χ
Figure B.1: Projection of the sensitive y-axis along Earth’s rotation axis provides the mag-nitude of response.
magnitude of Earth’s rotation response is given by the overlap of the y-direction and
Earth’s rotation axis (Fig. B.1). From Eq. 3.43, we expect the response to be
Ω⊕γn
(1− Qγn
γe
)cosχ ' 271 fT (B.1)
in effective magnetic units where Ω⊕ = 2π/(86164.09 s) for one sidereal day. This
is the amplitude of a sine wave with clear maximum along the N-S axis of the lab
(Fig. 3.16).
Early measurements showed the phase shifted 20-30 W and the response ampli-
tudes 20-30% larger than predicted. A systematic study of the location and orien-
tation of the lab was initiated to determine if these were real effects. Ultimately,
this observation would result in identification of the probe beam background effects
(Section. 4.1), rotational sensitivity to the x-direction from nonzero Bz and Lz fields
(Section. 4.4), and an overly precise determination in the location of the laboratory.
227
B.1.1 Jadwin Hall Exterior
An orthorectified aerial photograph (Fig. B.2) provides a useful measurement of Jad-
win Hall with respect to latitude and longitude. This image was taken as part of
the New Jersey state aerial survey1 conducted between February-April 2002. Image
G10A12 was taken from an altitude of 9600 feet above mean terrain and combined
with simultaneous GPS data to determine its geospatial location. The image has been
corrected from original photo negative to account for camera lens distortion, vertical
displacement, film shrinkage, atmospheric refraction, and variations in aircraft alti-
tude and orientation. Figure B.2 has an accuracy of ±4.0 feet at the 95% confidence
level. More details are available in the extensive metadata from the survey.
The latitude and longitude in Fig. B.2 are displayed in the North American Datum
of 1983 (NAD 83) New Jersey State Plane coordinate system. Because the Earth is
not a sphere, cartographers use a best fit ellipsoid. The NAD 83 geodetic reference
system is a datum, a system of reference, optimized for accurate representation of
North America and also serves as the legal horizontal coordinate system for the United
States government. The origin of the ellipsoid is 2 m from the center of the Earth
[121] and considered stable such that North America should not experience horizontal
motion with respect to NAD 83 coordinates. The State Plane coordinate system is
a developable surface, meaning that the curved surface is not stretched or torn when
represented on a flat plane. The New Jersey State plane uses the surface of a cylinder
with the axis along a central longitude line to project the state onto a flat plane. The
distortions are largest to the east and west far from the center longitude. Despite these
effects, the convergence of the longitude lines in New Jersey State Plane Coordinates
is in good agreement with the rotation axis of the Earth and can be confirmed using
a standard geographic information system (GIS) software tool, ArcGIS.
1Orthorectified images from state mapping surveys are freely available through the GeographicInformation Systems (GIS) section of the New Jersey Department of Environmental Protection(NJDEP) website.
228
74°39'4"W
74°39'4"W
74°39'5"W
74°39'5"W
74°39'5"W
74°39'5"W
74°39'6"W
74°39'6"W
74°39'7"W
74°39'7"W
74°39'7"W
74°39'7"W
74°39'7"W
74°39'7"W
74°39'8"W
74°39'8"W
74°39'8"W
74°39'8"W
74°39'9"W
74°39'9"W
74°39'9"W
74°39'9"W
74°39'10"W
74°39'10"W
74°39'10"W
74°39'10"W
74°39'11"W
74°39'11"W
74°39'11"W
74°39'11"W
40°20'46"N 40°20'46"N
40°20'46"N 40°20'46"N
40°20'45"N 40°20'45"N
40°20'45"N 40°20'45"N
40°20'44"N 40°20'44"N
40°20'44"N 40°20'44"N
40°20'44"N 40°20'44"N
40°20'43"N 40°20'43"N
40°20'43"N 40°20'43"N
40°20'42"N 40°20'42"N
40°20'41"N 40°20'41"N
40°20'41"N 40°20'41"N
40°20'40"N 40°20'40"N
40°20'40"N 40°20'40"N
40°20'40"N 40°20'40"N
Figure B.2: Aerial image of Jadwin Hall orthorectified to NAD83 State plane coordinates(2002).
229
A line drawn along the western edge of Jadwin Hall in Fig. B.2 determines a
heading of 29.5 W. Furthermore, a check of the latitude and longitude of two points
along this edge in ArcGIS gives a heading2 of 29.7 W. A similar result is achieved for
confirmation using Google MapsTM aerial photo. Google uses images set in the World
Geodetic System of 1984 (WGS 84) coordinate system. There are slight differences
between WGS 84 and NAD 83 in latitude and longitude, but the heading remains
the same to within our measurement precision. WGS 84 was developed by the US
Department of Defense specifically for global coverage rather than focusing on North
America.
We considered a commercial Global Positioning System (GPS) measurement along
the outside of the building. This type of measurement is well suited for measurements
where a clean line of sight is simultaneously available to at least four satellites. The
GPS system relies on precise timing of the microwave signal between the receiver and
satellites to accurately determine the distance. Any reflection off the building or con-
struction equipment generally littering the outside of area could bias any attempted
measurement. Also, an accurate determination of the heading depends on a long
baseline which is generally unavailable due to construction equipment surrounding
the building.
To circumvent these issues, we asked the Princeton University Facilities Depart-
ment to independently investigate the orientation of Jadwin Hall (Fig. B.3). Using
known benchmarks around the campus, Facilities Engineers surveyed the western
edge of the building using a theodolite. The hash marks indicate points that were
surveyed. This method determines an orientation of 30.35 W of Grid North. Grid
North is subtly different from True North3. True North is towards the North Pole as
determined by the Earth’s rotation axis. Grid North is determined by the Universal
2One should be careful to determine the distances on the surface of the sphere since latitude andlongitude are a measure of angle not distance.
3There is a third “North” of interest, Magnetic North. See section B.1.4.
230
Figure B.3: Princeton University Facilities Survey of Jadwin Hall. The hash marks indicatesurvey points. Grid North in this area is 12’ E of True North where we identify the NorthPole.
231
Transverse Mercator (UTM) grid. In the vicinity of Princeton, NJ, Grid North is
12’ E of True North implying that survey indicates that Jadwin lies 30.15 E of True
North. The 0.6 discrepancy with the aerial photos is smaller than uncertainties with
construction of the CPT-II equipment rack.
B.1.2 Jadwin Hall Interior
Figure B.4 shows the basement floorplan of Jawdin Hall where the experiment is
located in room B17. The north arrow in the lower right gives a heading of 35
in disagreement with the aerial photos. It is impossible to determine how the civil
engineer determined this angle. For a map such as this, it is a common practice for
surveyors to give only a general sense of a northerly direction. Also, surveyors will
often lay out a “Local North” arbitrarily for the job site. Therefore, we reject this
arrow for any precise determination of direction. We assume that the interior walls
of Jadwin are parallel to the exterior walls. We could request a survey of the lab with
reference to a known outdoor benchmark much like surveying an underground mine.
We lack motivation to embark on this project and choose to assume parallel interior
and exterior walls.
B.1.3 Comagnetometer Orientation
The axes of the comagnetometer are determined by the laser light fields. The sensi-
tive direction is a vector mutually orthogonal to propagation of the pump and probe
beams. One could determine this direction by removing optics and projecting the
probe beam across the room. This is most useful after optimization of comagnetome-
ter noise before taking data; however, it is at this point that the experimenter is
most loath to remove any optics. The propagation direction of the pump and probe
beams are constrained by the placement of the vacuum tubes entering the shields.
It is difficult to precisely measure the propagation direction, though centering the
232
Figure B.4: Floor plan of the basement of Jadwin Hall at Princeton University. Theexperiment is located in room B17.
233
beam to within 1 mm on both the windows gives alignment within 0.2. This is
much smaller than the absolute uncertainty in the placement of the windows with
respect to the optical breadboard, the orientation of the floating vibration isolation
platform, and the construction of the equipment rack. The absolute orientation of
all of these pieces can be determined with progressively smaller uncertainties for in-
creasing measurement effort. However, the experiments described in this thesis rely
on relative changes with respect to the comagnetometer, so these measurements were
not pursued.
In addition to the alignment of the light fields with the frame, it is likely the
magnetic field coils are not precisely lined up with the with the rest of the apparatus.
Mixing of field orientations with respect to the comagnetometer axes can cause small
offsets in the various zeroing routines as a modulation of By can become a modulation
of By+θBx where θ is the misalignment angle. Our control software allows for mixing
of one magnetic field basis into the light-field basis of up to a 5 misalignment. We
have not found a significant discrepancy to warrant an investigation of these effects.
B.1.4 Magnetic North
Magnetic North is irrelevant in determining the response of the magnetometer. The
comagnetometer inherits its sensitivity from a SERF magnetometer, but is not sen-
sitive to magnetic fields. The self-compensation and magnetic shields reduce the
influence of Earth’s magnetic field to 0.5 fT. Furthermore, the large Helmholtz coils
surrounding the experiment further suppress Earth’s magnetic field two orders of
magnitude. The magnetic North Pole lies in northern Canada and is distinct from
the Earth’s rotation axis. The magnetic declination, the difference between True
North and Magnetic North, in Princeton, New Jersey is 12.6 W of True North in
2010. Due to magnetic pole drift, it drifts at 1’ E per year.
234
Command Center
Jadwin Hall B17
VacuumPump
MotionControl
Electronics
4 kW Motor
C-clamp Jacks
Large 3-Axis Magnetic Field Coils
HomePosition
RotationPlatform
True North
Magnetic North
29.5°
12.6°
Figure B.5: Overhead view of the CPT-II apparatus in room B17.
235
B.2 Coordinate Frame Variations
This discussion so far has determined the spatial relationship between the experiment
and a fixed axis. It is interesting to consider the changes in Earth’s rotation vector
because it is known to vary. There are three degrees of freedom; the axial component
corresponds to the rate of rotation while the transverse components correspond to
polar motion. Most civil works and military construction projects do not require
reference to polar drift or tectonic plate motion [121], therefore, we expect these effects
to be small. These variations are not captured in the NAD 83 datum referenced in the
preceding sections, but are included in the International Terrestrial Reference Frame
(ITRF). The origin of the ITRF is at the center of the Earth and it is continuously
updated by Very Long Baseline Interferometry (VLBI) measurements. Several intense
VLBI campaigns have captured high resolution changes to the Earth’s rotation vector
[122], so these parameters are well-known.
B.2.1 Polar Motion
Polar motion is often reported in units of milliarcseconds (mas) of wobble. The
comagnetometer is most sensitive to polar motion when oriented along the E-W axis.
For a 1 aT response of the comagnetometer, the required polar motion would be
1000 mas or a 30 m shift of the pole. The largest measured oscillations are at the
level of 100 mas (Fig. B.6). These are a combination of a forced annual wobble and
the Chandler Wobble [123, 124]. Any irregularly shaped solid body rotating about an
axis misaligned with its figure axis, axis of symmetry, will exhibit Chandler Wobble.
The Earth has a measured Chandler Wobble period of 433 days that is continuously
driven by a combination of atmospheric and oceanic pressure fluctuations. The tides
produce diurnal and semidiurnal variations in the polar motion on a timescale relevant
to Lorentz violation searches; however, the largest of these are 0.5 milliarcseconds
236
Polar Drift
Pol
ar D
rift
(m
icro
arcs
econ
ds)
Time (days)
Time (days)
Len
gth o
f D
ay V
ari
atio
n (
µs)
Figure B.6: Processes affecting Earth’s rotation rate and polar drift. Adapted from [122].
peak-peak [125]. The long timescales and small scale of polar motion preclude any
relevant impact on the comagnetometer signal.
B.2.2 Length of Day
Regular revolutions of the Earth have long been used as a means of keeping time
referred to as Universal Time (UT). A correction to account for polar motion (Sec-
237
tion B.2.1) gives the standard UT1 found in the literature [125]. The variations in
UT1 are monitored by comparing rotations of the Earth against Ephemeris Time
following the motion of the solar system or Atomic Time kept by atomic clocks. It
seems confusing, but the literature will sometimes report rotation rate changes in
the length-of-day (LOD). For a 1 aT change in the comagnetometer signal, the LOD
would have to change by 0.3 s. The largest observed LOD changes are from El Nino
and the Southern Oscillation (ENSO) (Fig. B.6). These changes have a magnitude
500 µs and occur on timescales of years [126]. The tides also influence the rotation
of the Earth through conservation of angular momentum of the total Earth-ocean
system [127]. Variations in the length of day from the tides are on the order of 100 µs
peak-peak [125].
B.2.3 Earthquakes
On February 27, 2010, a magnitude 8.8 earthquake struck in Chile while the experi-
ment was running on sidereal day 3721. This event was large enough to modify both
Earth’s rotation rate and its axis of rotation. This is a consequence of conservation
of angular momentum under a modification of Earth’s inertia tensor. Based on the
1.26 µs change in the period of one day and the 2.7 milliarcsecond shift in the rotation
axis reported in the Jet Propulsion Laboratory press release [128], we would require
sensitivity of 3 zT/√
Hz to observe this effect. It is no surprise that no change in our
signal was observed.
B.2.4 Comment
Variations in Earth’s rotation do not influence the state of the art comagnetometer
performance and are not expected to in the near future. The comagnetometer’s
rotational sensitivity extends to an inertial frame, so it is interesting to consider
rotations beyond the Earth? In defining Earth’s rotation rate as the inverse of a
238
sidereal day, we already account for the Earth’s rotation about the Sun. But what
about the Sun’s rotation about our galaxy? This has a period of approximately 225
million terrestrial years [129] called a cosmic year, an exceptionally small rotation.
Larger scale, more sensitive ring laser gyroscopes have just begun to observe polar
motion from tides [130], and can safely constrain other possible rotations from an
inertial frame. It is interesting to note that the observed “tilt-over mode” (K1) in
Earth’s polar motion has a period of precisely one sidereal day. If the comagnetometer
ever reaches this level of sensitivity, the tilt-over mode will be major systematic in
future Lorentz violation searches.
B.3 Rotation Conventions
Positive rotations of the platform correspond to “spin up” under the right hand rule
convention. Almost all of the data has been taken with rotations in the positive
direction. We identify the “home” or zero degree position for the rotary encoder
when the edge of the platform holding the vertically mounted SRS lockin amplifier is
lined up by eye with the baseplate (Fig. B.5). The lock-in amplifier is a convenient
reference to a side of the apparatus carrying the y-axis of the comagnetometer. We
assume that the baseplate is perpendicular to the leg facing the command center.
The angle of this leg can be consistently measured with respect to the floor tiles. The
floor tiles seem to be reasonably well aligned with the interior walls of room B17. The
starting south position is typically a +(16− 18) rotation from the home position.
Over several weeks, the legs tend to rotate ∼ 1. This is because the entire
apparatus rests on the floor. We have chosen not to secure the legs to the foundation
of the building. In the event of an emergency stop, the shaft locks, and we have
observed the entire apparatus to spin up to 10 from a hard stop. If the equipment
were secured to the floor, this energy would be absorbed by the gears and shaft,
239
possibly damaging a critical piece of the rotation mechanism. We choose instead to
keep the area around the experiment clear of debris and dissipate the energy through
friction with the floor.
240
Appendix C
Comagnetometer Toy Model
There is a disconnect between the descriptive picture presented in Section 3.1.1 and
the steady state expansion of the Bloch equations presented in Section 3.2. It is not
immediately clear that the two are related. This appendix calculates the response of
an alkali magnetometer to an applied magnetic field if a compensating spin species
is present to cancel this field. This is a toy model treatment of the comagnetometer
behavior. While this treatment is heuristic, it is helpful to visualize how the nuclear
spins compensate for magnetic field changes from the geometry. Hopefully, this pre-
sentation will provide intuition to the reader as well as a meaningful insight to the
question “Why is this term in the Bloch equations solution?”.
C.1 Steady State Compensation
The geometrical approach to the comagnetometer behavior ignores spin-exchange
effects, pumping rates, and the finite magnetization of the electron spins. The treat-
ment in Section 3.2 can quantify these effects but requires a tour de force of algebraic
manipulation. The following treatment will not capture all of the fine details, but
provides a simple road to the general structure of the solution.
241
δBz Bca
δB+
ΔBz
Bca
ΔB+
Bn
θ
Figure C.1: The 3He follows the total field around the compensation point. This leaves theuncompensated fields ∆Bz and ∆B⊥.
C.1.1 Derivation
We decouple the electron and nuclear spins to consider an alkali-metal magnetometer
superimposed on top of the hyper-polarized noble-gas and enforce that the noble-
gas spins follow an adiabatic shift in the total field. We consider the response from
the reaction of the magnetometer to an incomplete cancelation of the applied field.
The magnetometer response to an applied fields is given by Eq. 2.48 and we find it
convenient to regroup the constants such that
P ex =
P ez γeRtot
[By + γe
RtotBxBz
1 + ( γeRtot
)2(B2x +B2
y +B2z )
]. (C.1)
Here, we have the framework for the numerator and denominator described earlier
where we can expand the denominator for Bi Rtot/γe = 20 µgauss. It is necessary
to determine which field the magnetometer will react to.
It is simplest to consider imperfect compensation by the noble-gas spins in two
dimensions. We consider an applied deviation from the compensation point as δBz
and δB⊥ (Fig. C.1). The 3He aligns to this direction, and the uncompensated field is
242
given by
∆B2dz = Ba
c + δBz −Bac cos θ (C.2)
∆B2d⊥ = δB⊥ −Ba
c sin θ (C.3)
where the approximation −Bn ≈ Bac since |Be| |Bn|. Both sin θ and cos θ can be
rewritten in terms of fields through the geometry of Fig. C.1 as
∆B2dz = Ba
c + δBz −Bac + δBz√
δB2⊥ + (1 + δBz)2
(C.4)
∆B2d⊥ = δB⊥ −
δB⊥√δB2⊥ + (Ba
c + δBz)2. (C.5)
The applied deviation δBi is reduced at first order, and the remnant shift of ∆Bi
depends on δB⊥, δBz, and Bac . The geometry is mildly more complicated in three
dimensions where these equations simplify to
∆Bx = Bx −Bx√
B2x +B2
y + (Bac +Bz)2
(C.6)
∆By = By −By√
B2x +B2
y + (Bac +Bz)2
(C.7)
∆Bz = (Bac +Bz)−
(Bac +Bz)√
B2x +B2
y + (Bac +Bz)2
(C.8)
where it has become convenient to drop the δBi notation and refer to the deviation
from the compensation point of a field Bi. We would ultimately like to include the
effects of anomalous fields and for shorthand introduce the definitions
~Be ≡~βe + ~L−~ΩQ(P e)
γe(C.9)
~Bn ≡~βn −~Ω
γn(C.10)
243
for generalized anomalous electron and nuclear couplings respectively. The distinction
between electron and nuclear couplings to rotations is distinguished by the appropri-
ate gyromagnetic ratio. We introduce the appropriate anomalous field to Eq. C.1 and
Eqs. C.6-C.8 as follows
P ex =
P ez γeRtot
[By + Bey + γe
Rtot(Bx + Bex)(Bz + Bez)
1 + ( γeRtot
)2[(Bx + Bex)2 + (By + Bey)2 + (Bz + Bez)2
]] (C.11)
∆Bx = Bx −Bx + Bnx√
(Bx + Bnx)2 + (By + Bny )2 + (Bac +Bz + Bnz )2
(C.12)
∆By = By −By + Bny√
(Bx + Bnx)2 + (By + Bny )2 + (Bac +Bz + Bnz )2
(C.13)
∆Bz = (Bac +Bz)−
(Bac +Bz + Bnz )√
(Bx + Bnx)2 + (By + Bny )2 + (Bac +Bz + Bnz )2
(C.14)
where we consider anomalous fields to influence the correction to the fields and not
the definition of the compensation point, Bac = −Bn.
Replacement of ∆Bi in Eqs. C.12-C.14 for Bi in Eq. C.11 easily provides a familiar
leading order response
P ex =
P eγeRtot
(βey − βny + Ly + Ωy
( 1
γn− Q(P e)
γe
))(C.15)
in the difference between electron and nuclear couplings. We can already start to
see why our Eq. 3.43 has a ByBz term and BxB2z from expansion of Eqs. C.12-C.14.
Further expansion of this expression to arbitrary orders can start to produce terms
listed in Section 3.2. For completeness of the argument, we expand all the terms
to an arbitrary third order product in fields to show the correlation between these
expressions and those produced earlier by a different means.
244
C.1.2 Steady State Equation Lists
The following sections provide lists of all the terms from the expansion in the geo-
metric method. We denote each of these as T (X) for “toy model” where X is the
variable of interest. This should distinguish the responses from our signal S(X) which
includes spin-exchange effects, pumping rates, and finite electron magnetization from
the Bloch equations. Terms have been ranked from large to small according to the
parameters listed in Table 3.2 and the modifications in Section 3.2.9 to maintain
terms for all components of the ~βe and ~βn fields.
Short of ranking terms, the lists constitute raw output. Since limits on ~βe and
~βn are already stringent, products of these components have been disregarded. Simi-
larly, T (X) considers coefficients of terms proportional to X, but does not explicitly
include terms proportional to X2. We therefore produce dependencies on products of
magnetic fields to complete some of the higher order picture. As before, we cannot
simply add all the dependencies together since this would double count some terms
and not others.
Algebraically grouping terms as in Section 3.2 has not been taken at this point.
This author has found the output in raw form very helpful to directly view throughout
the course of the work in this thesis. It seems useful to look through the list and
compare the relevant suppression of terms by γe/Rtot or Bc. Many of the terms beyond
the leading terms have so far been too small to observe, but in estimating their effects,
this author has gained a lot of intuition. Circling and assembling terms in these lists
has can develop intuition beyond the neatly presented1 effects in Chapter 3. These
lists also provide a sanity check in carefully expanding the denominator in Section 3.2.
All denominators have been expanded, so we must be careful when comparing T (X)
1Here, we present the complete list of terms from the expansion up to third order. Yes, it’s amess, but I believe it deserves to be reproduced in an appendix.
245
and S(X). They also enable the estimation of terms under different comagnetometer
conditions without sorting through a complicated Mathematica notebook once again.
It is important to keep in mind that field expansions as a product with two distinct
parameters have been performed here, Rtot/γe ≈ 20 µgauss and Bc ≈ 3 mgauss. These
expressions only hold for the smallest field deviations where the product is much less
than unity.
C.1.3 Primary Magnetic Field Dependence
Expansions for the magnetic fields match the expressions presented earlier with the
prominent ByBz and BxB2z terms. There is no linear dependance on Bx here since
this effect originated with spin-exchange terms.
T (Bz) =P ez γeRtot
Bz
(γeLxRtot
− γeβnx
Rtot
+γeβ
ex
Rtot
+By
Bc
+γeΩx
γnRtot
− 2γ2eLyLzR2tot
+2γ2
eLzβny
R2tot
− 2γ2eLzβ
ey
R2tot
− 2γ2eLyβ
ez
R2tot
+βnyBc
+BxγeLzBcRtot
− 2γ2eLzΩy
γnR2tot
− Ωy
Bcγn− QΩx
Rtot
+Bxγeβ
nz
BcRtot
+Bxγeβ
ez
BcRtot
− 2γ2eΩyβ
ez
γnR2tot
− BxγeΩz
BcγnRtot
+γeLzβ
nx
BcRtot
+2QγeLzΩy
R2tot
+2QγeLyΩz
R2tot
− 2Byβnz
B2c
− γeLzΩx
BcγnRtot
+2QγeΩyβ
ez
R2tot
− 2QγeΩzβny
R2tot
+2QγeΩzβ
ey
R2tot
+2ByΩz
B2cγn
− γeΩzβnx
BcγnRtot
− QBxΩz
BcRtot
− γeΩxβnz
BcγnRtot
− γeΩxβez
BcγnRtot
+2QγeΩyΩz
γnR2tot
+γeΩxΩz
Bcγ2nRtot
+2Ωyβ
nz
B2cγn
+2Ωzβ
ny
B2cγn
− QΩzβnx
BcRtot
− 2ΩyΩz
B2cγ
2n
− 2Q2ΩyΩz
R2tot
+QΩxΩz
BcγnRtot
)(C.16)
246
T (By) =P ez γeRtot
By
(Bz
Bc
+βnzBc
− Ωz
Bcγn− B2
z
B2c
+B2x
2B2c
+γeLxβ
ny
BcRtot
− γeLxΩy
BcγnRtot
− 2Bzβnz
B2c
+Bxβ
nx
B2c
+2BzΩz
B2cγn
+γeΩyβ
nx
BcγnRtot
− γeΩyβex
BcγnRtot
+γeΩxβ
ny
BcγnRtot
− BxΩx
B2cγn− γeΩxΩy
Bcγ2nRtot
− 3Ωyβny
B2cγn
+2Ωzβ
nz
B2cγn
− Ωxβnx
B2cγn− QΩxβ
ny
BcRtot
+3Ω2
y
2B2cγ
2n
− Ω2z
B2cγ
2n
+QΩxΩy
BcγnRtot
+Ω2x
2B2cγ
2n
)(C.17)
T (Bx) =P ez γeRtot
Bx
( B2zγe
BcRtot
+BzγeLzBcRtot
+Bzγeβ
nz
BcRtot
+Bzγeβ
ez
BcRtot
− BzγeΩz
BcγnRtot
+γeLzβ
nz
BcRtot
+γeLxβ
nx
BcRtot
− γeLzΩz
BcγnRtot
− γeLxΩx
BcγnRtot
+Byβ
nx
B2c
+2γeΩxβ
nx
BcγnRtot
− γeΩzβez
BcγnRtot
− QBzΩz
BcRtot
− γeΩxβex
BcγnRtot
− ByΩx
B2cγn− γeΩ
2x
Bcγ2nRtot
− Ωyβnx
B2cγn− QΩzβ
nz
BcRtot
− Ωxβny
B2cγn− QΩxβ
nx
BcRtot
+QΩ2
z
BcγnRtot
+ΩxΩy
B2cγ
2n
+QΩ2
x
BcγnRtot
)(C.18)
C.1.4 Products of Magnetic Fields
Combinations of two magnetic fields capture the higher order terms missed in Sec-
tion C.1.3. As expected, B2z has a Bx dependence.
T (ByBz) =P ez γeRtot
ByBz
Bc
(1− 2βnz
Bc
+2Ωz
Bcγn
)(C.19)
T (B2z ) =
P ez γeRtot
γeB2z
Rtot
(− γeLyRtot
+γeβ
ny
Rtot
− γeβey
Rtot
+Bx
Bc
− γeΩy
γnRtot
+βnxBc
− ByRtot
B2cγe
+QΩy
Rtot
− Ωx
Bcγn− Rtotβ
ny
B2cγe
+RtotΩy
B2cγeγn
)(C.20)
T (BxBy) =P ez γeRtot
BxBy
Bc
(βnxBc
− Ωx
Bcγn
)(C.21)
247
T (B2y) =
P ez γeRtot
γeB2y
Rtot
( Lx2Bc
− βnx2Bc
+βex
2Bc
+Ωx
2Bcγn+
3Rtotβny
2B2cγe− 3RtotΩy
2B2cγeγn
− QΩx
2Bcγe
)(C.22)
T (B2x) =
P ez γeRtot
γeB2x
Rtot
( Lx2Bc
− βnx2Bc
+βex
2Bc
+ByRtot
2B2cγe
+Ωx
2Bcγn+Rtotβ
ny
2B2cγe− RtotΩy
2B2cγeγn
− QΩx
2Bcγe
)(C.23)
T (BxBz) =P ez γeRtot
BxBz
Bc
(γeLzRtot
+γeβ
nz
Rtot
+γeβ
ez
Rtot
− γeΩz
γnRtot
− QΩz
Rtot
)(C.24)
248
C.1.5 Rotations
Rotation terms are in agreement with expressions in Section 3.2.7 after expansion of
the denominator.
T (Ωy) =P ez γeRtot
Ωy
γn
(1− Qγn
γe− B2
zγ2e
R2tot
− 2BzLzγ2e
R2tot
− Bz
Bc
− 3L2yγ
2e
R2tot
− 2βezBzγ2e
R2tot
− L2zγ
2e
R2tot
− L2xγ
2e
R2tot
+QB2
zγeγnR2tot
+6βnyLyγ
2e
R2tot
− 6βeyLyγ2e
R2tot
− 2βezLzγ2e
R2tot
+2βnxLxγ
2e
R2tot
− 2βexLxγ2e
R2tot
− βnzBc
+2QBzLzγeγn
R2tot
− ByLxγeBcRtot
+Ωz
Bcγn− 2Lxγ
2eΩx
R2totγn
+3QL2
yγeγn
R2tot
+2QβezBzγeγn
R2tot
− 3B2y
2B2c
+βnxByγeBcRtot
− βexByγeBcRtot
+QL2
zγeγnR2tot
+QL2
xγeγnR2tot
+B2z
B2c
− 6QβnyLyγeγn
R2tot
+6QβeyLyγeγn
R2tot
+2QBzγeΩz
R2tot
+2βnxγ
2eΩx
R2totγn
− 2βexγ2eΩx
R2totγn
− B2x
2B2c
+2QβezLzγeγn
R2tot
− 2QβnxLxγeγnR2tot
+2QβexLxγeγn
R2tot
− ByγeΩx
BcRtotγn− βnyLxγe
BcRtot
+4QLxγeΩx
R2tot
+2QLzγeΩz
R2tot
− γ2eΩ
2x
R2totγ
2n
− 3βnyBy
B2c
+2βnzBz
B2c
− βnxBx
B2c
− 4QβnxγeΩx
R2tot
+4QβexγeΩx
R2tot
+2QβezγeΩz
R2tot
− 2BzΩz
B2cγn
− 2Q2BzγnΩz
R2tot
− βny γeΩx
BcRtotγn+BxΩx
B2cγn
+QByΩx
BcRtot
+3QγeΩ
2x
R2totγn
− 2Q2LzγnΩz
R2tot
− 2Q2LxγnΩx
R2tot
− 2βnz Ωz
B2cγn
− 2Q2βezγnΩz
R2tot
+2Q2βnxγnΩx
R2tot
− 2Q2βexγnΩx
R2tot
+βnxΩx
B2cγn
+QβnyΩx
BcRtot
+Ω2z
B2cγ
2n
− Q2Ω2z
R2tot
− 3Q2Ω2x
R2tot
− Ω2x
2B2cγ
2n
+Q3γnΩ2
z
R2totγe
+Q3γnΩ2
x
R2totγe
)(C.25)
249
T (Ωx) =P ez γeRtot
Ωx
γn
(BzγeRtot
+LzγeRtot
+βezγeRtot
− QBzγnRtot
− 2LxLyγ2e
R2tot
− QLzγnRtot
− B2zγe
BcRtot
+B2yγe
2BcRtot
+B2xγe
2BcRtot
+2βnxLyγ
2e
R2tot
− 2βexLyγ2e
R2tot
+2βnyLxγ
2e
R2tot
− 2βeyLxγ2e
R2tot
− QβezγnRtot
− BzLzγeBcRtot
− BxLxγeBcRtot
− 2Lxγ2eΩy
R2totγn
− QΩz
Rtot
+2βnxBxγeBcRtot
+2QLxLyγeγn
R2tot
− βnzBzγeBcRtot
− βezBzγeBcRtot
+βnyByγe
BcRtot
− βexBxγeBcRtot
+2βnxγ
2eΩy
R2totγn
− 2βexγ2eΩy
R2totγn
− BxBy
B2c
− QB2yγn
2BcRtot
− QB2xγn
2BcRtot
− ByγeΩy
BcRtotγn
+BzγeΩz
BcRtotγn− 2QβnxLyγeγn
R2tot
+2QβexLyγeγn
R2tot
− 2QβnyLxγeγn
R2tot
+2QβeyLxγeγn
R2tot
− βnzLzγeBcRtot
− βnxLxγeBcRtot
+4QLxγeΩy
R2tot
+Q2γnΩz
Rtotγe+
LzγeΩz
BcRtotγn
− 4QβnxγeΩy
R2tot
+4QβexγeΩy
R2tot
− βnxBy
B2c
− βnyBx
B2c
− QβnyByγn
BcRtot
− QβnxBxγnBcRtot
− βny γeΩy
BcRtotγn+
βezγeΩz
BcRtotγn+BxΩy
B2cγn
+QByΩy
BcRtot
+QBzΩz
BcRtot
+γeΩ
2y
2BcRtotγ2n
− 2Q2LxγnΩy
R2tot
+2Q2βnxγnΩy
R2tot
− 2Q2βexγnΩy
R2tot
+βnxΩy
B2cγn
+QβnyΩy
BcRtot
+Qβnz Ωz
BcRtot
− QΩ2z
BcRtotγn− QΩ2
y
2BcRtotγn
)(C.26)
T (Ωz) =P ez γeRtot
Ωz
γn
(− By
Bc
− QLxγnRtot
− BxBzγeBcRtot
− βnyBc
+Qγnβ
nx
Rtot
− Qγnβex
Rtot
+2QBzγeLyγn
R2tot
− BxγeLzBcRtot
+Ωy
Bcγn− QΩx
Rtot
+2QγeLyLzγn
R2tot
− 2QBzγeγnβny
R2tot
+2QBzγeγnβ
ey
R2tot
+2ByBz
B2c
− Bzγeβnx
BcRtot
− Bxγeβez
BcRtot
− QBxBzγnBcRtot
+2QBzγeΩy
R2tot
− 2QγeLzγnβny
R2tot
+2QγeLzγnβ
ey
R2tot
+2QγeLyγnβ
ez
R2tot
+BzγeΩx
BcγnRtot
− γeLzβnx
BcRtot
+2QγeLzΩy
R2tot
+Q2γnΩx
γeRtot
+2Bzβ
ny
B2c
+2Byβ
nz
B2c
+γeLzΩx
BcγnRtot
− QBzγnβnx
BcRtot
− QBxγnβnz
BcRtot
+2QγeΩyβ
ez
R2tot
− 2BzΩy
B2cγn
− 2Q2BzγnΩy
R2tot
+γeΩxβ
ez
BcγnRtot
+QBzΩx
BcRtot
− 2Q2LzγnΩy
R2tot
− 2Ωyβnz
B2cγn
− 2Q2γnΩyβez
R2tot
+QΩxβ
nz
BcRtot
)(C.27)
250
C.1.6 Lightshifts
The LxLz term appears as well as the first order dependance of Ly.
T (Lz) =P ez γeRtot
Lz
(γeLxRtot
− γeβnx
Rtot
+γeβ
ex
Rtot
− 2Bzγ2eLy
R2tot
+γeΩx
γnRtot
+2Bzγ
2eβ
ny
R2tot
− 2Bzγ2eβ
ey
R2tot
+BxBzγeBcRtot
− 2Bzγ2eΩy
γnR2tot
− 2γ2eLyβ
ez
R2tot
− QΩx
Rtot
+Bzγeβ
nx
BcRtot
+Bxγeβ
nz
BcRtot
− 2γ2eΩyβ
ez
γnR2tot
+2QBzγeΩy
R2tot
− BxγeΩz
BcγnRtot
− BzγeΩx
BcγnRtot
+2QγeLyΩz
R2tot
+2QγeΩyβ
ez
R2tot
− 2QγeΩzβny
R2tot
+2QγeΩzβ
ey
R2tot
− γeΩzβnx
BcγnRtot
− γeΩxβnz
BcγnRtot
+2QγeΩyΩz
γnR2tot
+γeΩxΩz
Bcγ2nRtot
− 2Q2ΩyΩz
R2tot
)(C.28)
T (Lx) =P ez γeRtot
Lx
(BzγeRtot
+γeLzRtot
+γeβ
ez
Rtot
+B2yγe
2BcRtot
+B2xγe
2BcRtot
+2γ2
eLyβnx
R2tot
− 2γ2eLyβ
ex
R2tot
− QΩz
Rtot
− 2γ2eLyΩx
γnR2tot
+Byγeβ
ny
BcRtot
+Bxγeβ
nx
BcRtot
+2γ2
eΩyβnx
γnR2tot
− 2γ2eΩyβ
ex
γnR2tot
+2γ2
eΩxβny
γnR2tot
− 2γ2eΩxβ
ey
γnR2tot
− ByγeΩy
BcγnRtot
− BxγeΩx
BcγnRtot
− 2γ2eΩxΩy
γ2nR
2tot
+2QγeLyΩx
R2tot
− 2QγeΩyβnx
R2tot
+2QγeΩyβ
ex
R2tot
− γeΩyβny
BcγnRtot
− 2QγeΩxβny
R2tot
+2QγeΩxβ
ey
R2tot
− γeΩxβnx
BcγnRtot
+4QγeΩxΩy
γnR2tot
+γeΩ
2y
2Bcγ2nRtot
+γeΩ
2x
2Bcγ2nRtot
− 2Q2ΩxΩy
R2tot
)(C.29)
251
T (Ly) =P ez γeRtot
Ly
(1− B2
zγ2e
R2tot
− 2Bzγ2eLz
R2tot
− 2Bzγ2eβ
ez
R2tot
− γ2eL
2z
R2tot
− γ2eL
2x
R2tot
− 2γ2eLzβ
ez
R2tot
+2γ2
eLxβnx
R2tot
− 2γ2eLxβ
ex
R2tot
− 2γ2eLxΩx
γnR2tot
+6γ2
eΩyβny
γnR2tot
− 6γ2eΩyβ
ey
γnR2tot
+2QBzγeΩz
R2tot
+2γ2
eΩxβnx
γnR2tot
− 2γ2eΩxβ
ex
γnR2tot
− 3γ2eΩ
2y
γ2nR
2tot
+2QγeLzΩz
R2tot
− γ2eΩ
2x
γ2nR
2tot
+2QγeLxΩx
R2tot
− 6QγeΩyβny
R2tot
+6QγeΩyβ
ey
R2tot
+2QγeΩzβ
ez
R2tot
+6QγeΩ
2y
γnR2tot
− 2QγeΩxβnx
R2tot
+2QγeΩxβ
ex
R2tot
+2QγeΩ
2x
γnR2tot
− 3Q2Ω2y
R2tot
− Q2Ω2z
R2tot
− Q2Ω2x
R2tot
)(C.30)
C.1.7 Anomalous Fields
Expansions for the anomalous fields are in agreement with Sections 3.2.4 and 3.2.9.
T (βey) =P ez γeRtot
βey
(1− B2
zγ2e
R2tot
− 2Bzγ2eLz
R2tot
− 3γ2eL
2y
R2tot
− γ2eL
2z
R2tot
− γ2eL
2x
R2tot
− 6γ2eLyΩy
γnR2tot
− 2γ2eLxΩx
γnR2tot
+2QBzγeΩz
R2tot
− 3γ2eΩ
2y
γ2nR
2tot
+6QγeLyΩy
R2tot
+2QγeLzΩz
R2tot
− γ2eΩ
2x
γ2nR
2tot
+2QγeLxΩx
R2tot
+6QγeΩ
2y
γnR2tot
+2QγeΩ
2x
γnR2tot
− 3Q2Ω2y
R2tot
− Q2Ω2z
R2tot
− Q2Ω2x
R2tot
)(C.31)
252
T (βny ) =P ez γeRtot
βny
(− 1 +
B2zγ
2e
R2tot
+2Bzγ
2eLz
R2tot
+Bz
Bc
+3γ2
eL2y
R2tot
+γ2eL
2z
R2tot
+γ2eL
2x
R2tot
+6γ2
eLyΩy
γnR2tot
+ByγeLxBcRtot
− Ωz
Bcγn+
2γ2eLxΩx
γnR2tot
+3B2
y
2B2c
− B2z
B2c
− 2QBzγeΩz
R2tot
+B2x
2B2c
+3γ2
eΩ2y
γ2nR
2tot
− 6QγeLyΩy
R2tot
+ByγeΩx
BcγnRtot
− 2QγeLzΩz
R2tot
− γeLxΩy
BcγnRtot
+γ2eΩ
2x
γ2nR
2tot
− 2QγeLxΩx
R2tot
− 3ByΩy
B2cγn
− 6QγeΩ2y
γnR2tot
+2BzΩz
B2cγn
− BxΩx
B2cγn
− QByΩx
BcRtot
− γeΩxΩy
Bcγ2nRtot
− 2QγeΩ2x
γnR2tot
+3Q2Ω2
y
R2tot
+3Ω2
y
2B2cγ
2n
− Ω2z
B2cγ
2n
+Q2Ω2
z
R2tot
+QΩxΩy
BcγnRtot
+Q2Ω2
x
R2tot
+Ω2x
2B2cγ
2n
)(C.32)
T (βex) =P ez γeRtot
βex
(BzγeRtot
+γeLzRtot
− 2γ2eLxLyR2tot
+B2yγe
2BcRtot
+B2xγe
2BcRtot
− 2γ2eLxΩy
γnR2tot
− QΩz
Rtot
− 2γ2eLyΩx
γnR2tot
− ByγeΩy
BcγnRtot
− BxγeΩx
BcγnRtot
− 2γ2eΩxΩy
γ2nR
2tot
+2QγeLxΩy
R2tot
+2QγeLyΩx
R2tot
+4QγeΩxΩy
γnR2tot
+γeΩ
2y
2Bcγ2nRtot
+γeΩ
2x
2Bcγ2nRtot
− 2Q2ΩxΩy
R2tot
)(C.33)
T (βnx ) =P ez γeRtot
βnx
(− Bzγe
Rtot
− γeLzRtot
+2γ2
eLxLyR2tot
+B2zγe
BcRtot
− B2yγe
2BcRtot
− B2xγe
2BcRtot
+BzγeLzBcRtot
+BxγeLxBcRtot
+2γ2
eLxΩy
γnR2tot
+QΩz
Rtot
+2γ2
eLyΩx
γnR2tot
+BxBy
B2c
+ByγeΩy
BcγnRtot
+2BxγeΩx
BcγnRtot
− BzγeΩz
BcγnRtot
+2γ2
eΩxΩy
γ2nR
2tot
− 2QγeLxΩy
R2tot
− 2QγeLyΩx
R2tot
− γeLzΩz
BcγnRtot
− γeLxΩx
BcγnRtot
− BxΩy
B2cγn− QBzΩz
BcRtot
− 4QγeΩxΩy
γnR2tot
− ByΩx
B2cγn− QBxΩx
BcRtot
− γeΩ2y
2Bcγ2nRtot
− 3γeΩ2x
2Bcγ2nRtot
+2Q2ΩxΩy
R2tot
+QΩ2
z
BcγnRtot
+ΩxΩy
B2cγ
2n
+QΩ2
x
BcγnRtot
)(C.34)
253
T (βez) =P ez γeRtot
βez
(γeLxRtot
− 2Bzγ2eLy
R2tot
+γeΩx
γnRtot
− 2γ2eLyLzR2tot
+BxBzγeBcRtot
− 2Bzγ2eΩy
γnR2tot
− 2γ2eLzΩy
γnR2tot
− QΩx
Rtot
+2QBzγeΩy
R2tot
− BxγeΩz
BcγnRtot
− BzγeΩx
BcγnRtot
+2QγeLzΩy
R2tot
+2QγeLyΩz
R2tot
+2QγeΩyΩz
γnR2tot
+γeΩxΩz
Bcγ2nRtot
− 2Q2ΩyΩz
R2tot
)(C.35)
T (βnz ) =P ez γeRtot
βnz
(By
Bc
+BxBzγeBcRtot
+BxγeLzBcRtot
− Ωy
Bcγn− 2ByBz
B2c
− BzγeΩx
BcγnRtot
− γeLzΩx
BcγnRtot
+2BzΩy
B2cγn
+2ByΩz
B2cγn
− QBxΩz
BcRtot
− 2ΩyΩz
B2cγ
2n
+QΩxΩz
BcγnRtot
)(C.36)
C.1.8 Signal
Sorting through expressions Eq. C.16-C.36 to select the leading terms produces,
TL =P ez γeRtot
[βey − βny +
Ωy
γn− ΩyQ
γe+ Ly +Bz
(By + βnyBc
+ LxγeRtot
+γeΩx
Rtotγn
)+
γeRtot
(BxBz(Bz + Lz)
Bc
− LyγeRtot
B2z −
γeΩy
RtotγnB2z + LxLz
)](C.37)
which is precisely the same as Eq. 3.43 produced by a solution to the complete Bloch
equations that disregards pumping rates and a finite electron magnetization.
254
Appendix D
Lorentz Violation Summary
This appendix provides details on each of the data sets that comprises the 143 days
of data collected for the Lorentz violation search in Chapter 5. Each of the files is
labeled by the GMST value near the start of the run. The tables are broken in two
halves to distinguish between the summer and winter collections.
Table D.1 provides information regarding the selection of data in each record. The
winter collection implemented the probe background measurements (Section 4.1.7)
referred to as Secondary data. The effective noise bandwidth (ENBW) is calculated
by comparing the average spectrum over 60 selections and follows the procedure
described in Section E.2. A butterworth filter with a 2 Hz cuttoff, 5 Hz stopband, 3 dB
passband ripple, and 60 dB stopband attenuation precedes the ENBW calculation to
remove oscillations at the compensation frequency. It is curious to note that the
effective noise bandwidth is larger for the background measurements than for the
comagnetometer signal.
Table D.2 provides details on the rate of sampling the comagnetometer signal and
the lock-in bandwidth on these measurements. Notes regarding significant changes to
the operation of the apparatus or interrupted operation are listed. Datasets 3541.74-
3553.04 and 3669.09-3685.43 were taken sequentially with only a brief pause to create
255
a new file and resume operation. The data is collected at 200 Hz and averaged down
to the frequency reported as the sample rate. There is a minor technical issue in
analyzing the data where the entire dataset must be loaded into on contiguous block
of memory. For data saved at 40 Hz, this means that analysis is simplest if a new file
is created every week. This aligns well with a regular performance check on a weekly
basis.
Table D.3 provides the time duration of each file as well as the particular rotation
scheme applied. Most of the rotations occur with 22 s intervals except for file 3720.51.
Likewise, rotations in this file occur at the “slow” rotation speed. All files except file
3553.04 start in the south location and the direction of rotations was reversed at
file 3754.58. These changes were made to check against lingering systematic effects;
however, not enough data was collected for a realistic comparison. The orientation of
the 3He spin provides an important such check. The extra rotation column describes
the switching from the N-S to E-W axes. For example, file 3536.96 starts in the
“south” position and rotates 8 times with an extra rotation of +90. This implies
that the positions are [south, north, south, north, south, north, south, north, west]
before the next minor zeroing event. The minor zeroing events follow the ordering
[south, west, north, east, south,...].
The length of the files in the winter are significantly longer than those in the sum-
mer. This is due to a computer upgrade in the fall of 2009. The original computer
did not have enough system resources to average the signal to low sample rates. In
addition, this computer would experience intermittent DAQ errors that would halt
operation when the hardware lost timing. Simply clicking “OK” on the screen would
resume normal operation. These errors were intermittent enough that vigilance in
monitoring the experiment was required as it was unclear at the time what caused
the errors. Small gaps exist in the data when these issues were identified right away.
Larger gaps initiated the start of a new file. These problems where eventually identi-
256
fied as a flaw in the PCI bridge allowing two PCI cards to be used an in a single PCI
slot on the compact computer. A new computer for second half of the data set with
two PCI slots, more memory, and faster CPU solved this issue that has not returned.
257
Primary SecondaryFile Window Offset ENBW Window Offset ENBW
Number (s) (s) (Hz) (s) (s) (Hz)3492.86 3 0 1.7 - - -3495.44 3 0 1.7 - - -3498.72 3 0 1.7 - - -3500.97 3 0 1.7 - - -3504.05 3 0 1.7 - - -3506.63 10 0 1.7 - - -3510.80 7 0 - - - -3511.53 7 0 1.7 - - -3513.63 7 0 1.7 - - -3518.31 7 0 1.7 - - -3519.56 7 0 1.7 - - -3524.54 7 0 1.7 - - -3536.96 7 0 1.7 - - -3541.74 7 0 1.78 - - -3544.79 7 0 1.78 - - -3548.96 7 0 1.78 - - -3553.04 7 0 1.78 - - -3651.18 3 -5 1.5 2.0 0 23656.54 3 -5 2 2.0 0 23663.19 3 -5 1.5 2.0 0 4.53666.47 3 -5 1.25 2.0 0 43669.09 3 -5 1.5 2.0 0 43677.13 3 -5 1.5 2.0 0 43685.43 3 -5 1.5 2.0 0 43692.34 3 -5 1.5 2.0 0 33711.32 3 -5.3 1.5 2.0 -0.3 63720.51 3 -5 2.0 2.0 0 83734.60 3 -5 2 2.0 0 83741.32 3 -5 2.12 2.0 0 8.953754.58 3 -5 2 2.0 0 83756.33 3 -5 2 2.0 0 8
Table D.1: Secondary data was implemented for the second half of data collection. File3510.80 has no ENBW since it is dropped from the data set.
258
File Sample Lock-inNote
Number Rate (s) τ (ms)3492.86 130 3 Noisy traditional calibration improved3495.44 130 3 DAQ Error3498.72 130 33500.97 130 33504.05 130 3 Early drifts3506.63 80 3 Gap in data3510.80 80 3 Bad file3511.53 80 3 Lock-in knob rotates to second harmonic3513.63 80 3 Giorgos repairs with duct tape and restarts3518.31 80 3 Mike start with large pump wavelength offset3519.56 80 3 Virus scanner interrups; large one day oscillations3524.54 80 3 Major zeroing to 7 hour intervals; E-W bifurcation3536.96 80 3 2 hour gap; E-W discontinuity3541.74 80 3 (1/4) Wavemeter fails, linear interpolation corrects3544.79 80 3 (2/4)3548.96 80 3 (3/4)3553.04 80 3 (4/4)
3651.18 40 100Christmas run; major zeroing remotely from MA;multiple wavemeter and encoder failures
3656.54 40 100 Mike start3663.19 40 100 Reduce medium data rate from 10 Hz to 2.5 Hz3666.47 40 100 Computer crash; three records without fast data3669.09 40 100 (1/3) Princeton OIT turns off internet3677.13 40 100 (2/3)3685.43 40 100 (3/3) Unknown PC reboot terminates longest run3692.34 40 100 HBz in zeroing; all ByBz slopes starting at 3693.38
3711.32 40 100Tilt sensor 3; spectrum degraded; wavemeter crash;adjust triggers from clock synchronization error
3720.51 40 100Slow speed; wavemeter failure at start; stress platedisabled; rest measurements; major zero south only
3734.60 50 100Loose contact causes heater drift; polarization driftgives large E-W error bars; wavemeter gaps
3741.32 40 100 Stem connectors repair3754.58 40 100 Reverse rotations; temperature oscillations increase3756.33 40 100 Stabilized temperature
Table D.2: Listing of datasets from the Lorentz violation search with notes on relevantevents.
259
File Duration Interval Interval Start Rot. ExtraSpeed
3HeNumber (sd) Number Delay (s) Loc. Dir. Rot. Spin3492.86 2.53 16 22 south up 90 mod. up3495.44 3.11 10 22 south up 90 mod. up3498.72 2.15 10 22 south up 90 mod. up3500.97 3.01 10 22 south up 90 mod. up3504.05 2.08 10 22 south up 90 mod. up3506.63 3.78 10 22 south up 90 mod. up3510.80 0.60 10 22 south up 90 mod. up3511.53 1.89 10 22 south up 90 mod. up3513.63 4.57 10 22 south up 90 mod. up3518.31 1.18 10 22 south up 90 mod. up3519.56 4.59 10 22 south up 90 mod. up3524.54 5.07 10 22 south up 90 mod. up3536.96 4.70 8 22 south up 90 mod. up3541.74 2.99 8 22 south up 90 mod. up3544.79 4.16 8 22 south up 90 mod. up3548.96 4.06 8 22 south up 90 mod. up3553.04 1.62 8 22 north up 90 mod. up3651.18 5.29 8 22 south up 90 mod. dn3656.54 6.61 8 22 south up 90 mod. dn3663.19 3.14 8 22 south up 90 mod. dn3666.47 2.32 8 22 south up 90 mod. dn3669.09 8.01 8 22 south up 90 mod. dn3677.13 8.30 8 22 south up 90 mod. dn3685.43 5.50 8 22 south up 90 mod. dn3692.34 8.65 8 22 south up 90 mod. dn3711.32 9.14 8 22 south up 90 mod. up3720.51 9.59 8 25 south up 90 slow up3734.60 6.50 8 22 south up 90 mod. up3741.32 10.86 8 22 south up 90 mod. up3754.58 1.69 8 22 south dn -90 mod. up3756.33 6.28 8 22 south dn -90 mod. up
Table D.3: Summary of parameters relevant to the rotation of the apparatus.
260
Appendix E
Signal Analysis
The results presented in this thesis depend on a careful data analysis and understand-
ing the uncertainty in the measurement. Many of these ideas are well-known and can
be referenced from a variety of resources [131]. The technical details contributing to
the measurements in this thesis are described here.
E.1 Uncertainty
In a real measurement, not all recorded values are precisely the same and are subject
to random variations. The size of these random variations leads to the statistical
uncertainty in the measurement. This section outlines the general methods for deter-
mining statistical uncertainty. In what follows, the systematic uncertainty, a bias in
the measurement is neglected. There is a clear prescription for treatment of statis-
tical uncertainties whereas treatment of systematic uncertainty is handled on a case
by case basis.
261
E.1.1 Independent Measurements
Given N independent, equally-weighted measurements of a quantity x subject to
random variations, the best estimate of x is the central value or mean
x =N∑i=1
xiN
(E.1)
where xi is the value of each measurement. The variance of this distribution is
σ2x =
N∑i=1
(x− xi)2
N − 1(E.2)
where the approximation N − 1 ≈ N is common when N is large. We can represent
the uncertainty in the measurement as the standard deviation of the mean
σx =σx√N
(E.3)
which can been seen to decrease as the square root of the number of measurements.
σx is sometimes referred to as the standard error or standard error of the mean.
E.1.2 Weighted Average and Uncertainty
If there is uncertainty associated with each measurement, the average of these quan-
tities should reflect the relative precision of the contributing values. Given N inde-
pendent measurements of the same quantity xi with associated uncertainties σi, the
weighted average of the xi measurements is defined as
〈x〉 =
∑Ni=1wixi∑Ni=1 wi
(E.4)
262
where the weights are defined as wi = 1/σ2i . The uncertainty in the weighted average
is
σ〈x〉 =1√∑Ni=1 wi
. (E.5)
The values with the smallest σi carry the largest statistical weight.
E.1.3 Reduced χ2
The uncertainty σi in measurement xi often represents the best estimate of the un-
certainty in the measurement. It is important to check the relationship between the
reported σi and the Gaussian distribution of xi that we expect. The quantity χ2
defined as
χ2 =N∑i=1
(〈x〉 − xi)2
σ2i
(E.6)
is useful in performing this check. It is often more standard to consider the reduced
chi-squared
χ2 = χ2/d (E.7)
where d is the number of degrees of freedom given by N − c where c is the number of
constraints in the model. A χ2 that is close to 1 is often a good check that the error
bars are representative of the scatter. If χ2 6' 1, this indicates that the scatter in the
data are larger than the representative error bars. Similarly, if χ2 < 1, the scatter in
the data is smaller than the representative error bars.
If χ2 is not close to 1, reinvestigation of how the initial σi are determined is
recommended because this typically indicates a more appropriate estimate is required.
In some cases, it is reasonable to readjust σi by the χ2 of the distribution such that
σ′i = σi√χ2. (E.8)
263
If experimental parameters are changing over the timescale of many measurements but
are relatively stable over an individual measurement, χ2 will be large. Correcting by
the reduced chi-squared will capture this variation, though it may be more worthwhile
to stabilize these changing parameters if they are known. Caution is recommended
here as this is not always the most appropriate representation of the uncertainty
and this step should be carefully justified. A more careful estimate of the initial
uncertainties can result in a better estimate of the uncertainty.
E.1.4 Error Propagation
A parameter of interest is often a function of several variables. If we are interested in
the uncertainty in function q(x, y, ...z) and measurements of x± δx, y ± δy,...z ± δz
are available, then
δq(x, y, ...z) =
√(∂q∂xδx)2
+(∂q∂yδy)2
+ ...(∂q∂zδz)2
. (E.9)
In the case that q = x + y, this expression generalizes to the well-known result of
adding uncertainties in quadrature
δq =√
(δx)2 + (δy)2. (E.10)
Similarly, the case of q = xy generates the expression
δq
q=
√(δxx
)2
+(δyy
)2
(E.11)
which is the sum of the fractional uncertainties in quadrature.
264
E.2 Over-Sampled Measurements
The definition of the standard deviation of the mean in Eq. E.3 relies on the as-
sumption that each measured point is independent. In recording analog signals, the
sample frequency is at least twice the bandwidth to avoid aliasing. If the bandwidth
is f , then a signal recorded at 2f or 10f should have the same uncertainty; over-
sampling the data faster than the correlations does not decrease the uncertainty. The
effective noise bandwidth1 (ENBW) of the filter will correct for over-sampling if the
characteristics of the filter are known. Equation E.3 becomes
σx =σx√
N × ENBW/(fs/2)(E.12)
where fs is the sample rate. The number of independent points is modified by the
ratio of the effective noise bandwidth to the Nyquist frequency, fs/2.
The comagnetometer measurements in this thesis involves a SR830 lock-in ampli-
fier. The predicted ENBW for this device with the slope of the low pass filter set to
24 dB/octave is 5/(64τ) where τ is the lock-in time constant [132]. Practically, the
lock-in signal is digitally recorded from the analog output of the lock-in by a data
acquisition card. Higher frequency components beyond the lock-in bandwidth are
reintroduced to the signal in the form of white noise from analog to digital conver-
sion. Therefore, it is important to be able to determine the effective noise bandwidth
of an arbitrary signal2.
The ENBW is defined as the bandwidth of an ideal brick filter that passes the
same power as the non-ideal filter. By taking the fourier transform of the data in the
1Sometimes called the noise bandwidth.2This point can be circumvented by digitally recording the data from the lock-in, however a
fast and robust bit rate along with synchronization to other analog channels introduces some othertechnical hurdles.
265
0 2 4 6 8 10 12 14 16 18 200
0.1
0.2
0.3
0.4
0.5
0.6
|A(f
k)|2
(μV
/Hz)
2
Frequency (Hz)
ENBW
Figure E.1: Determination of the effective noise bandwidth of 2.12 Hz for file 3741.32.
time domain, this can be written as
ENBW =1
|A(0)|2∫ ∞
0
|A(f)|2df (E.13)
where A(f) is the fourier component of the signal at frequency f [133]. This operation
turns an arbitrary spectrum into that of the idealized brick case with a constant and
a sharp cut-off in the power spectrum. The effective noise bandwidth is the frequency
of the cut-off (Fig. E.1) and predicts the value of 5/(64τ) for the 24 dB/octave low
pass filter on the lock-in. In the case of a discrete signal, Eq. E.13 becomes
ENBW =
∑Nk=0 |A(fk)|2N |A(f0)|2
fs2
(E.14)
where N = fs/(2T )+1, T is the measurement interval, and k is the discrete frequency
component in the fourier transform.
E.3 String Analysis
In the experiments described in this thesis, we search for changes in the signal with
respect to a binary change in either the spin source or apparatus orientation. We
could simply compare the difference in pairs of points in either state to determine the
266
correlated amplitude. This method is highly susceptible to drifts in the signal over
many reversals. Instead, we follow the string analysis techniques first introduced in a
neutron electric dipole moment search [134]. This technique removes systematic shifts
in measurements of the reversal correlated amplitude. We exclusively use a 3-point
overlapping string analysis and discuss those details. Generalization to higher order
strings is available in [135].
Consider a sequence of regularly sampled measurements subject to a linear drift
where the parameter of interest a is reversed
yi = x0 + ai
yi+1 = x0 − ai + bt
yi+2 = x0 + ai + 2bt
....
where yi is the ith measurement, x0 is the starting offset, and b is the slope of linear
drift in time t. We assume that the measurements are short compared to the time
between measurements such that the drift does not significantly affect the uncer-
tainty σi in measurement yi. Given these assumptions, the following combination of
measurements
ai =(−1)i+1
4(yi − 2yi+1 + yi+2) (E.15)
form a string which removes the linear drift. This amplitude can be calculated for
additional consecutive measurements. The uncertainty in ai is given by
σai =1
4
√σ2i + 4σ2
i+1 + σi+2. (E.16)
These uncertainties are independent and add as in Eq. E.10. The coefficient on the
middle term is important since this element has been counted twice. Naturally, this
267
technique can be extended to arbitrary order and written in compact form using
binomial coefficients to remove higher order drifts [135].
Each string represents a measurement of a. The selection of a particular string
is arbitrary, so we consider all possible 3-point strings and allow them to overlap.
Typically, one also finds better statistical uncertainty using overlapping strings. The
weighted average of the strings over a measurement interval can be expressed as
〈a〉 =
∑N−2i=1 ai/(σai)
2∑N−2i=1 1/(σai)
2(E.17)
where N is the number of measurements. The uncertainty of 〈a〉 is given by
σ〈a〉 =fo√∑N−2
i=1 1/(σai)2
√χ2 (E.18)
where fo = 4/√
6 is a statistical weight correction for the 3-point overlapping strings.
The reduced χ2 is defined much like in Section E.1.3 as
χ2 =1
N − 2
L−2∑i=1
(ai − 〈a〉)2
σ2ai
. (E.19)
Explicitly rescaling Eq. E.18 by√χ2 serves to help match the uncertainty to the
scatter in the string points. This technique removes a linear drift the in the yi mea-
surements and over short intervals can suppress quadratic drift. It is straightforward
to generalize this technique to higher order strings to remove higher order drifts [135].
E.3.1 String Analysis Refinement
The string analysis described in Section E.3 works very well for long strings N 10
such as those in an EDM experiment [135]. Due to the zeroing schedule of the
comagnetometer, measurement intervals are limited to ∼ 250 s limiting 8-16 reversals
268
in the case of apparatus rotations in CPT-II. We refine the string analysis to more
accurately represent measurements in the low reversal limit.
We reject the correction by the square root of the χ2 parameter. This value is
often less than 1 and serves to shrink the uncertainty. This is not from a flaw in our
data. Even for white gaussian noise, it takes up to ∼ 30 measured points for the
distribution of χ2 to be peaked near 1 as expected. In addition, measurements with
N ∼ 10 do not provide strong enough statistics for this correction to be representative
of the measurement interval.
We also need to correct the mysterious fo parameter for short strings. This cor-
rectly accounts for uneven weighting of the string points in the string analysis to
what one might get from a naive ordering of the points. In the naive sense, one would
compute the differences in successive measurements and find the weighted average
ignoring any linear drift
〈an〉 =
∑N/2j=1(y2j−1 − y2j)/(σ
22j−1 + σ2
2j)
2∑N/2
j=1 1/(σ22j−1 + σ2
2j)(E.20)
where N ∈ 2Z for simplicity in comparing pairs of points with the corresponding
uncertainty
σ〈an〉 =1√∑N/2
j=1 4/(σ22j−1 + σ2
2j)=
σo√N
(E.21)
where the last step assumes that each σi = σo. This is the statistical precision to which
we know 〈an〉. Since we have the same amount of information but chose to group
it differently in the calculation of 〈a〉, the uncertainty should be the same. Clever
grouping in the string analysis removes a systematic bias of linear drift and should
retain the same number of degrees of freedom. In determining σ〈a〉 and ignoring the
269
adjustment by√χ2,
σ〈a〉 =fo√∑N−2
i=1 ( 42
σ2i +4σ2
i+1+σ2i+2
)=fo√
6
4
σo√N − 2
(E.22)
where again in the last step we assume σi = σo. We clearly observe the appearance
of 4/√
6. A comparison of Eq. E.21 and E.22 indicates that
fo =4√6
√N − 2
N(E.23)
is a more appropriate assignment. In the limit N 1,√
1− 2/N ≈ 1 and is ignored
in Ref. [135]. In the case of N = 10 common in the Lorentz violation experiment, this
leads to a 10% correction in the uncertainty which is significant in the propagation
of uncertainty. We could wonder if the above argument is somewhat oversimplified.
Starting with Eq. E.17 and using Eq. E.11, the result simplifies to the above argument
for overlapping strings.
The modification to fo just described was not applied in the analysis of the Lorentz
violation experiment described in Chapter 5. When the least squares fit to sidereal
amplitudes is applied, the uncertainties in the fit parameters are rescaled by the√χ2
of the fit. Since the modification is the same throughout the entire file, this overall
scaling factor isn’t necessary. If different length measurement intervals were used
in a file, this correction would be necessary to restore the relative weights of the
uncertainties.
270
Bibliography
[1] Scott Jeffery Seltzer. Developements in Alkali-Metal Atomic Magnetometry.PhD thesis, Princeton University, 2008.
[2] S. J. Seltzer, P. J. Meares, and M. V. Romalis. Synchronous optical pump-ing of quantum revival beats for atomic magnetometry. Phys. Rev. A,75(5):051407(R), 2007.
[3] Ronald J. Adler. Six easy roads to the Planck scale. Am. J. Phys., 78(9), 2010.
[4] C. Rovelli. Loop quantum gravity. Living Rev. Relativity, 1, 1998.
[5] Amit Giveon and David Kutasov. Brane dynamics and gauge theory. Rev. Mod.Phys., 71(4), 1999.
[6] R. J. Adler, P. Chen, and D. I. Santiago. The generalized uncertainty principleand black hole remnants. Gen. Relativ. Gravit., 33, 2001.
[7] J. Ellis, N. E. Mavromatos, and D. V. Nanopoulos. Derivation of a vacuumrefractive index in a stringy space-time foam model. Phys. Lett. B, 665, 2008.
[8] Joseph Silk. The cosmic microwave background. Annales Henri Poincare,4:275–290, 2003.
[9] D. Colladay and V. A. Kostelecky. Lorentz-violating extension of the standardmodel. Phys. Rev. D, 58(11):116002, 1998.
[10] Don Colladay and V. Alan Kostelecky. CPT violation and the standard model.Phys. Rev. D, 55(11), 1997.
[11] V. Alan Kostelecky and Stuart Samuel. Spontaneous breaking of Lorentz sym-metry in string theory. Phys. Rev. D, 39(2), 1989.
[12] O. W. Greenberg. CPT violation implies violation of Lorentz invariance. Phys.Rev. Lett., 89(23), 2002.
[13] Achison and Hey. Gauge Theories in Particle Physics. Institute of Physics,2003.
[14] J. J. Sakurai and Jim Napolitano. Modern Quantum Mechanics. Addison-Wesley, 2011.
271
[15] G. Luders. On the equaivalence of invariance under time reversal and underparticle–anti-particle conjugation for relativistic field theories. K. Dan Vidensk.Selsk. Mat.-fys. Medd., 28(5), 1954.
[16] W. Pauli. Exclusion Principle, Lorentz Group and Reflection of space-time andcharge. Pergamon Press, New York, 1955.
[17] V. Alan Koskelecky and Charles D. Lane. Constraints on Lorentz violationfrom clock-comparison experiments. Phys. Rev. D, 60(11):116010, 1999.
[18] C. S. Wu, E. Ambler, R. W. Hayward, D. D. Hoppes, and R. P. Hudson.Experimental test of parity conservation in beta decay. Phys. Rev., 105(4),1957.
[19] J. H. Christenson, J. W. Cronin, V. L. Fitch, and R. Turlay. Evidence for the2π decay of the K0
2 meson. Phys. Rev. Lett., 13(4), 1964.
[20] K. Tsigutkin, D. Dounas-Frazer, A. Family, J. E. Stalnaker, V. V. Yashchuk,and D. Budker. Observation of a large atomic parity violation effect in Ytter-bium. Phys. Rev. Lett., 103(7), 2009.
[21] S. C. Bennett and C. E. Wienman. Measurement of the 6S → 7S transitionpolarizability in atomic Cesium and an improved test of the standard model.Phys. Rev. Lett., 82(12), 1999.
[22] O. Bertolami, Don Colladay, V. Alan Kostelecky, and R. Potting. CPT violationand baryogenesis. Physics Letters B, 395, 1997.
[23] C. Q. Geng and Lei Geng. Some tests on CPT invariance. Modern Phyiscs A,22(7-10):651–659, 2007.
[24] V. W. Hughes, H. G. Robinson, and V. Beltran-Lopez. Upper limit for theanisotropy of inertial mass from nuclear resonance experiments. Phys. Rev.Lett., 4(7), 1960.
[25] R. W. P. Drever. A search for anisotropy of inertial mass using a free precessiontechnique. Philosophical Magazine, 6(65):683–687, 1961.
[26] V. Alan Kostelecky. Topics in Lorentz and CPT violation. In V. AlanKostelecky, editor, Proceedings of the Fifth Meeting on CPT and Lorentz Sym-metry. World Scientific, 2011.
[27] V. Alan Kostelecky and Neil Russell. Data tables for Lorentz and CPT violation.Rev. Mod. Phys., 83, 2011.
[28] Thomas Whitmore Kornack. A Test of CPT and Lorentz Symmetry Using aK-3He Comagnetometer. PhD thesis, Princeton University, 2005.
272
[29] D. Bear, R. E. Stoner, R. L. Walsworth, V. Alan Kostelecky, and Charles D.Lane. Limit on Lorentz and CPT violation of the neutron using a two-speciesnoble-gas maser. Phys. Rev. Lett., 85(24), 2000.
[30] J.M. Brown, S.J. Smullin, T.W. Kornack, and M.V. Romalis. New limiton Lorentz- and CPT-violating neutron spin interactions. Phys. Rev. Lett.,105(14):151604, 2010.
[31] J. D. Prestage, J. J. Bollinger, Wayne M. Itano, and D. J. Wineland. Limitsfor spatial anisotropy by use of nuclear-spin–polarized 9Be+ ions. Phys. Rev.Lett., 54(22), 1985.
[32] S. K. Lamoreaux, J. P. Jacobs, B. R. Heckel, F. J. Raab, and E. N. Fortson.New limits on spatial anisotropy from optically pumped 201Hg and 199Hg. Phys.Rev. Lett., 57(25), 1986.
[33] T. E. Chupp, R. J. Hoare, R. A. Loveman, E. R. Oteiza, J. M. Richardson, andM. E. Wagshul. Results of a new test of local Lorentz invariance: A search formass anisotropy in 21Ne. Phys. Rev. Lett., 63(15), 1989.
[34] C. J. Berglund, L. R. Hunter, Jr. D. Krause, M. S. Ronfelt E. O. Prigge, andS. K. Lamoreaux. New limits on local Lorentz invariance from Hg and Csmagnetometers. Phys. Rev. Lett., 75(10), 1995.
[35] D. F. Phillips, M. A. Humphrey, E. M. Mattison, R. E. Stoner, R. C. F. Vessot,and R. L. Walsworth. Limit on Lorentz and CPT violation of the proton usinga hydrogen maser. Phys. Rev. D, 63(11):111101(R), 2001.
[36] D. Bear, R. E. Stoner, R. L. Walsworth, V. Alan Kostelecky, and Charles D.Lane. Erratum: Limit on Lorentz and CPT violation of the neutron using atwo-species noble-gas maser [Phys. Rev. Lett. 85, 5038 (2000)]. Phys. Rev. Lett.,89(20), 2002.
[37] F. Cane, D. Bear, D. F. Phillips, M. S. Rosen, C. L. Smallwood, R. E. Stoner,R. L. Walsworth, and V. A. Kostelecky. Bound on Lorentz- and CPT-violatingboost effects for the neutron. Phys. Rev. Lett., 93(23):230801, 2004.
[38] Peter Wolf, Frederic Chapelet, Sebastien Bize, and Andre Clairon. Coldatom clock test of Lorentz invariance in the matter sector. Phys. Rev. Lett.,96(6):060801, 2006.
[39] I. Altarev et.al. Test of Lorentz invariance with spin precession of ultracoldneutrons. Phys. Rev. Lett., 103(8):081602, 2009.
[40] C. Gemmel, W. Heil, S. Karpuk, K. Lenz, Yu. Sobolev, K. Tullney, M. Burghoff,W. Kilian, S. Knappe-Gruneberg, W. Muller, A. Schnabel, F. Seifert,L. Trahms, and U. Schmidt. Limit on Lorentz and CPT violation of thebound neutron using a free precession 3He/129Xe comagnetometer. Phys. Rev.D, 82(11):111901(R), 2010.
273
[41] T. W. Kornack, R. K. Ghosh, and M. V. Romalis. Nuclear spin gyroscope basedon an atomic magnetometer. Phys. Rev. Lett., 95(23):230801, 2005.
[42] G. Vasilakis, J.M. Brown, T.W. Kornack, and M.V. Romalis. Limits on newlong range nuclear spin-dependant forces set with a K-3He comagnetometer.Phys. Rev. Lett., 103(26):261801, 2009.
[43] Giorgos Vasilakis. Precision Measurements of Spin Interactions with High Den-sity Atomic Vapors. PhD thesis, Princeton University, 2011.
[44] Alan Corney. Atomic and Laser Spectroscopy. Clarendon Press, 1977.
[45] N. Lwin and D. G. McCartan. Collision broadening of the potassium resonancelines by noble gases. Journal of Physics B, 11(22), 1978.
[46] David R. Lide, editor. Handbook of Chemistry and Physics. CRC Press, 85edition, 2004.
[47] J. Ma, A. Kishinevski, Y.-Y. Jau, C. Reuter, and W. Happer. Modificiation ofglass cell walls by rubidium vapor. Phys. Rev. A, 79(4):042905, 2009.
[48] I. M. Savukov, S. J. Seltzer, M. V. Romalis, and K. L. Sauer. Tunable atomicmagnetometer for detection of radio-frequency magnetic fields. Phys. Rev. Lett.,95(6):063004, 2005.
[49] W. Happer and B. S. Mathur. Effective operator formalism in optical pumping.Physical Review, 163(1), 1967.
[50] S. Appelt, A. Ben-Amar Baranga, C. J. Erickson, M. V. Romalis, A. R. Young,and W. Happer. Theory of spin-exchange optical pumping of 3He and 129Xe.Phys. Rev. A, 58(2), 1998.
[51] William Happer. Optical pumping. Reiews of Modern Physics, 44(2), 1972.
[52] F. A. Franz and C. Volk. Electronic spin relaxation of the 4S21/2 state of K
induced by K-He and K-Ne collisions. Phys. Rev. A, 26(1), 1982.
[53] Thad G. Walker and William Happer. Spin-exchange optical pumping of noble-gas nuclei. Rev. Mod. Phys., 69(2), 1997.
[54] E. Babcock, B. Chann, T. G. Walker, W. C. Chen, and T. R. Gentile. Limitsto the polarization for spin-exchange optical pumping of 3He. Phys. Rev. Lett.,96(8):083003, 2006.
[55] N. R. Newbury, A. S. Barton, G. D. Cates, W. Happer, and H. Middleton.Gaseous 3He-3He magnetic dipolar spin relaxation. Phys. Rev. A, 48(6), 1993.
[56] L. D. Schearer and G. K. Walters. Nuclear spin-lattice relaxation in the presenceof magnetic-field gradients. Physical Review, 139(5a):1398, 1965.
274
[57] G. D. Cates, S. R. Schaefer, and W. Happer. Relaxation of spins due to fieldinhomogeneities in gaseous samples at low magnetic fields and low pressures.Phys. Rev. A, 37(8), 1988.
[58] J. Schmiedeskamp, W. Heil, E. W. Otten, R. K. Kremer, A. Simon, and J. Zim-mer. Paramagnetic relaxation of spin polarized 3He at bare glass surfaces. Eur.Phys. J. D, 38(3), 2006.
[59] A. Deninger, W. Heil, E. W. Otten, M. Wolf, R. K. Kremer, and A. Simon.Paramagnetic relaxation of spin polarized 3He at coated glass walls. Eur. Phys.J. D, 38(3), 2006.
[60] J. Schmiedeskamp, H.-J. Elmers, W. Heil, E. W. Otten, Yu. Sobolev, W. Kilian,H. Rinneberg, T. Sander-Thommes, F. Seifert, and J. Zimmer. Relaxation ofspin polarized 3He by magnetized ferromagnetic contaminants. Eur. Phys. J.D, 38(3), 2006.
[61] A. Abragam. The Principles of Nuclear Magnetism. Oxford University Press,1961.
[62] Dmitry Budker, Derek F. Kimball, and David P. DeMille. Atomic Physics: anexploration through problems and solutions. Oxford University Press, 2004.
[63] H. G. Dehmelt. Modulation of a light beam by precessing absorbing atoms.Physical Review, 105(6), 1957.
[64] William Bell and Arnold L. Bloom. Optical detection of magnetic resonance inalkali-metal vapor. Physical Review, 107(6), 1957.
[65] J. C. Allred, R. N. Lyman, T. W. Kornack, and M. V. Romalis. High-sensitivityatomic magnetometer unaffected by spin-exchange relaxation. Phys. Rev. Lett.,89(13), 2002.
[66] H. B. Dang, A. C. Maloof, and M. V. Romalis. Ultrahigh sensitivity magneticfield and magnetization measurements with an atomic magnetometer. AppliedPhysics Letters, 97(15):151110, 2010.
[67] W. Happer and H. Tang. Spin-exchange shift and narrowing of magnetic reso-nance lines in optically pumped alkali vapors. Phys. Rev. Lett., 31(5), 1973.
[68] W. Happer and A. C. Tam. Effect of rapid spin exchange on the magnetic-resonance spectrum of alkali vapors. Phys. Rev. A, 16(5), 1977.
[69] I. M. Savukov and M. V. Romalis. Effect of spin-exchange collisions in a high-density alkali-metal vapor in low magnetic fields. Phys. Rev. A, 71(2):023405,2005.
[70] T. W. Kornack and M. V. Romalis. Dynamics of two overlapping spin ensemblesinteracting by spin exchange. Phys. Rev. Lett., 89(25), 2002.
275
[71] S. R. Schaefer, G. D. Cates, Ting-Ray Chien, D. Gonatas, W. Happer, andT. G. Walker. Frequency shifts of the magnetic-resonance spectrum of mixturesof nuclear spin-polarized noble gases and vapors of spin-polarized alkali-metalatoms. Phys. Rev. A, 39(11), 1989.
[72] B. C. Grover. Noble-gas NMR detection through noble-gas-rubidium hyperfinecontact interaction. Phys. Rev. Lett., 40(391), 1978.
[73] Earl Babcock, Ian A. Nelson, Steve Kadlecek, and Thad G. Walker. 3Hepolarization-dependant EPR frequency shifts of alkali-metal-3He pairs. Phys.Rev. A, 71(1):013414, 2005.
[74] Lawrence W. Cheuk. Progress towards a K-Rb-21Ne comagnetometer. Under-graduate Thesis, Princeton University, 2010.
[75] Michael Romalis, Joel Allred, and Robert Lyman. K-3He self-compensatingcomagnetometer for tests of CPT symmetry. In V. Alan Kostelecky, editor,Proceedings of the Second Meeting on CPT and Lorentz Symmetry. World Sci-entific, 2002.
[76] T. W. Kornack and M. V. Romalis. Operation of the K-3He self-compensatingcomagnetometer for a test of Lorentz symmetry. In V. Alan Kostelecky, edi-tor, Proceedings of the Third Meeting on CPT and Lorentz Symmetry. WorldScientific, 2005.
[77] T. W. Kornack, G. Vasilakis, and M. V. Romalis. Preliminary results from atest of CPT and Lorentz symmetry using a K-3He comagnetometer. In V. AlanKostelecky, editor, Proceedings of the Fourth Meeting on CPT and LorentzSymmetry. World Scientific, 2008.
[78] T. W. Kornack, S. J. Smullin, S.-K. Lee, and M. V. Romalis. A low-noise ferritemagnetic shield. Applied Physics Letters, 90(22):223501, 2007.
[79] S. J. Smullin, T. W. Kornack, G. Vasilakis, and M. V. Romalis. A second-generation comagnetometer for testing fundamental symmetries. In V. AlanKostelecky, editor, Proceedings of the Fourth Meeting on CPT and LorentzSymmetry. World Scientific, 2008.
[80] Eugene Hecht, editor. Optics. Addison Wesley Longman, Inc., 1998.
[81] S. Ramaseshan, F.A.Sc., and V. Sivaramakrishnan. Faraday rotation in calcite.Proc. Ind. Acad. Sci. A, 44, 1956.
[82] Paul A. Williams, A. H. Rose, G. W. Day, T. E. Milner, and M. N. Deeter.Temperature dependence of the Verdet constant in several diamagnetic glasses.Applied Optics, 30(10), 1991.
[83] Charles C. Robinson. The Faraday rotation of diamagnetic glasses from 0.334µrad to 1.9 µrad. Applied Optics, 3(10), 1964.
276
[84] Richard Emil Jacob. Studies of Hyperpolarized 3He Relaxation at Glass Sur-faces. PhD thesis, The University of Utah, 2003.
[85] Jared Snow Teter. Wall Relaxation Studies of Hyperpolarized 3He. PhD thesis,The University of Utah, 2005.
[86] B. R. Heckel, E. G. Adelberger, C. E. Cramer, T. S. Cook, S. Schlamminger, andU. Schmidt. Preferred-frame and CP-violation tests with polarized electrons.Phys. Rev. D, 78(9):092006, 2008.
[87] Pavel A. Bolokhov, Stefan Groot Nibbelink, and Maxim Pospelov.Lorentz-violating supersymmetric quantum electrodynamics. Phys. Rev. D,72(1):015013, 2005.
[88] Pavel A. Bolokhov and Maxim Pospelov. Classification of dimension-5 Lorentz-violating interactions in the standard model. Phys. Rev. D, 77(2):025022, 2008.
[89] J. L. Friar, B. F. Gibson, G. L. Payne, A. M. Bernstein, and T. E. Chupp.Neutron polarization in polarized 3He targets. Phys. Rev. C, 42(6), 1990.
[90] Brett Altschul. Extending the reach of Lorentz tests with atomic clocks. InV. Alan Kostelecky, editor, Proceedings of the Fifth Meeting on CPT andLorentz Symmetry. World Scientific, 2011.
[91] V. Alan Kostelecky, Neil Russell, and Jay D. Tasson. Constraints on torsionfrom bounds on Lorentz violation. Phys. Rev. Lett., 100(11):111102, 2008.
[92] Victor Flambaum, Simon Lambert, and Maxim Pospelov. Scalar-tensor theorieswith pseudoscalar couplings. Phys. Rev. D, 80(10):105021, 2009.
[93] Robert C. Myers and Maxim Pospelov. Ultraviolet modifications of dispersionrelations in effective field theory. Phys. Rev. Lett., 90(21):211601, 2003.
[94] Irina Mocioiu, Maxim Pospelov, and Radu Roiban. Breaking CPT by mixednoncommutativity. Phys. Rev. D, 65(10):107702, 2002.
[95] Dominik J. Schwarz, Glenn D. Starkman, Dragan Huterer, and Craig J. Copi.Is the low-l microwave background cosmic? Phys. Rev. Lett., 93(22):221301,2004.
[96] M. V. Romalis. Hybrid optical pumping of optically dense alkali-metal vaporwithout quenching gas. Phys. Rev. Lett., 105(24):243001, 2010.
[97] M. Smiciklas, J. M. Brown, L. W. Cheuk, and M. V. Romalis. A new limiton Lorentz-violating tensor interactions set using 21Ne-Rb-K comagnetometer.arXiv:1106.0738v1, 2011.
[98] Alexander G. Glenday, Claire E. Cramer, David F. Phillips, and Ronald L.Walsworth. Limits on anomalous spin-spin couplings between neutrons. Phys.Rev. Lett., 101(26), 2008.
277
[99] X. Zheng et al. Precision measurement of the neutron spin asymmetries andspin-dependant structure functions in the valence quark region. Phys. Rev. C,70(6), 2007.
[100] T. R. Gentile, W. C. Chen, G. L. Jones, E. Babcock, and T. G. Walker. Polar-ized 3He spin filters for slow neutron physics. J. Res. Natl. Inst. Stand. Technol.,110(3):299–304, 2005.
[101] Justin M. Brown. A search for an anomalous spin-dependant force betweennuclear spins. Technical report, Princeton University, 2007.
[102] Mikhail V. Romalis. Laser Polarized 3He Target Used for a Precision Measure-ment of the Neutron Spin Structure. PhD thesis, Princeton University, 1997.
[103] W. C. Chen, T. R. Gentile, T. G. Walker, and E. Babcock. Spin-exchangeoptical pumping of 3He with Rb-K mixtures and pure K. Phys. Rev. A,75(1):013416, 2007.
[104] B. Chann, E. Babcock, L. W. Anderson, and T. G. Walker. Measurements of3He spin-exchange rates. Phys. Rev. A, 66(3):032703, 2002.
[105] Earl D. Babcock. Spin-Exchange Optical Pumping with Alkali-Metal Vapors.PhD thesis, University of Wisconsin-Madison, 2005.
[106] W. J. Mullin, F. Laloe, and M. G. Richards. Longitudinal relaxation time fordilute quantum gases. Journal of Low Temperature Physcs, 80(1/2), 1990.
[107] B. Lancor and T. G. Walker. Polarization limits in K-Rb spin-exchange mix-tures. Phys. Rev. A, 83(6):065401, 2011.
[108] Thad G. Walker. Private communication. May 2008.
[109] Nomogram of allowable pressures for borosilicate glass tubes. Kimble KontesTechnical Datasheet.
[110] M. V. Romalis and G. D. Cates. Accurate 3He polarimetery using the RbZeeman frequency shift due to the Rb-3He spin-exchange collisions. Phys. Rev.A, 58(4), 1998.
[111] Hao Xu, Steven M. Conolly, Greig C. Scott, and Albert Macovski. Homogeneousmagnet design using linear programming. IEEE Transactions on Magnetics,36(2), 2000.
[112] Milan Wayne Garrett. Thick cylindrical coil systems for strong magnetic fieldswith field or gradient homogeneities of the 6th to 20th order. Journal of AppliedPhysics, 38(6), 1967.
[113] Milan Wayne Garrett. Table of solenoids with sixth-order error and near-maximum power efficiency. Journal of Applied Physics, 40(8), 1969.
278
[114] J. P. Jacob. Search for a permanent electric dipole moment on Mercury 199atoms as a test of time reversal symmetry. PhD thesis, University of Washing-ton, 1991.
[115] E. Babcock, A. Petoukhov, J. Chastagnier, D. Jullien, E. Lelievre-Berna, K. H.Anderson, R. Georgii, S. Masalovich, S. Boag, C. D. Frost, and S. R. Parnell.AFP flipper devices: Polarized 3He spin flipper and shorter wavelength neutronflipper. Physica B: Condensed Matter, 397(1-2), 2007.
[116] John David Jackson. Classical Electrodynamics. John Wiley & Sons, Inc., thirdedition, 1999.
[117] B. A. Dobrescu. Massless gauge bosons other than the photon. Phys. Rev. Lett.,94(15), 2005.
[118] A. Bohr and B. R. Mottelson. Nuclear Structure, volume 1. Benjamin, NewYork, 1969.
[119] Nima Arkani-Hamed, Hsin-Chia Cheng, Markus Luty, and Jesse Thaler. Uni-versal dynamics of spontaneous Lorentz violation and a new spin-dependantinverse-square law force. Jounal of High Energy Physics, (07):029, 2005.
[120] Robin M. Green. Spherical Astronomy. Cambridge University Press, 1985.
[121] US Army Corps of Engineers. Engineering and Design - NAVSTAR GlobalPositioning System Surveying, 2003.
[122] T. A. Clark, C. Ma, J. W. Ryan, B. F. Chao, J. M. Gipson, D. S. MacMillan,N. R. Vandenberg, T. M. Eubanks, and A. E. Niell. Earth rotation measure-ments yields valuable information about the dynamics of the earth system. Eos,Transations, American Geophysical Union, 79(17), 1998.
[123] Richard S. Gross. The excitation of the Chandler wobble. Geophysical ResearchLetters, 27(15), 2000.
[124] John M. Gipson and Chopo Ma. Site displacement due to variation in Earthrotation. Journal of Geophysical Research, 103(B4), 1998.
[125] B. F. Chao and R. D. Ray. Oceanic tidal angular momentum and Earth’srotation variations. Progress in Oceanography, 40, 1997.
[126] Benjamin F. Chao. Geodesy is not just for static measurements any more. Eos,Transactions, American Geophysical Union, 84(16), 2003.
[127] Richard S. Gross. The effect of ocean tides on the Earth’s rotation as predictedby the results of an ocean tide model. Geophysical Research Letter, 20(4), 1993.
[128] Chilean quake may have shortened Earth days. NASA Jet Propulsion Labora-tory Press Release, March 2010.
279
[129] Jerry Jon Sellers, William J. Astore, Robert B. Giffen, and Wiley J. Larson.Understanding Space: An Introduction to Astronautics. McGraw-Hill, 2005.
[130] K. U. Schreiber, A. Velikoseltsev, M. Rothacher, T. Klugel, G. E. Stedman,and D. L. White. Direct measurement of diurnal polar motion by ring lasergyroscopes. Journal of Geophysical Research, 109(B06405), 2004.
[131] John R. Taylor. An Introduction to Error Analysis: The Study of Uncertaintiesin Physical Measurements. University Science Books, 1997.
[132] Stanford Research Systems, 1290-D Reamwood Avenue, Sunnyvale, CA 94089.Model SR830 DSP Lock-in Amplifier, 2.3 edition, 1993.
[133] Jacob Millman and Christos C. Halkias. Integrated Electronics: analog anddigital circuits and systems. McGraw-Hill Book Company, 1972.
[134] W. B. Dress, P. D. Miller, Paul Perrin Pendlebury, and Norman F. Ramsey.Search for an electric dipole moment of the neutron. Phys. Rev. D, 15(1), 1977.
[135] William Clark Griffith. Limiting CP Violation Through a Search for a Perma-nent Electric Dipole Moment of Mercury 199 Atoms. PhD thesis, University ofWashington, 2005.
280
top related